modelId
stringlengths 5
139
| author
stringlengths 2
42
| last_modified
timestamp[us, tz=UTC]date 2020-02-15 11:33:14
2025-06-27 00:42:13
| downloads
int64 0
223M
| likes
int64 0
11.7k
| library_name
stringclasses 499
values | tags
sequencelengths 1
4.05k
| pipeline_tag
stringclasses 54
values | createdAt
timestamp[us, tz=UTC]date 2022-03-02 23:29:04
2025-06-27 00:40:00
| card
stringlengths 11
1.01M
|
|---|---|---|---|---|---|---|---|---|---|
PrunaAI/pxyyy-rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa-bnb-smashed | PrunaAI | 2024-11-13T13:53:46Z | 5 | 0 | null | [
"safetensors",
"llama",
"pruna-ai",
"base_model:pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa",
"base_model:quantized:pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa",
"8-bit",
"bitsandbytes",
"region:us"
] | null | 2024-11-12T04:34:54Z | ---
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
base_model: pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
tags:
- pruna-ai
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed with llm-int8.
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We use safetensors.
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check requirements from the original repo pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa installed. In particular, check python, cuda, and transformers versions.
1. Make sure that you have installed quantization related packages.
```bash
pip install transformers accelerate bitsandbytes>0.37.0
```
2. Load & run the model.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("PrunaAI/pxyyy-rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa-bnb-smashed", trust_remote_code=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained("pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa")
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
outputs = model.generate(input_ids, max_new_tokens=216)
tokenizer.decode(outputs[0])
```
## Configurations
The configuration info are in `smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model pxyyy/rlhflow_mixture_clean_empty_round_with_dart_downsampled-600k-wlisa before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/SingularityHJY-GPTNeoX-160M-Minipile-bnb-smashed | PrunaAI | 2024-11-13T13:53:25Z | 5 | 0 | null | [
"safetensors",
"gpt_neox",
"pruna-ai",
"custom_code",
"base_model:SingularityHJY/GPTNeoX-160M-Minipile",
"base_model:quantized:SingularityHJY/GPTNeoX-160M-Minipile",
"8-bit",
"bitsandbytes",
"region:us"
] | null | 2024-11-11T23:08:34Z | ---
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
base_model: SingularityHJY/GPTNeoX-160M-Minipile
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
tags:
- pruna-ai
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed with llm-int8.
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We use safetensors.
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check requirements from the original repo SingularityHJY/GPTNeoX-160M-Minipile installed. In particular, check python, cuda, and transformers versions.
1. Make sure that you have installed quantization related packages.
```bash
pip install transformers accelerate bitsandbytes>0.37.0
```
2. Load & run the model.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("PrunaAI/SingularityHJY-GPTNeoX-160M-Minipile-bnb-smashed", trust_remote_code=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained("SingularityHJY/GPTNeoX-160M-Minipile")
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
outputs = model.generate(input_ids, max_new_tokens=216)
tokenizer.decode(outputs[0])
```
## Configurations
The configuration info are in `smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model SingularityHJY/GPTNeoX-160M-Minipile before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/lalainy-ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1-bnb-smashed | PrunaAI | 2024-11-13T13:53:23Z | 5 | 0 | null | [
"safetensors",
"qwen2",
"pruna-ai",
"base_model:lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1",
"base_model:quantized:lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1",
"8-bit",
"bitsandbytes",
"region:us"
] | null | 2024-11-12T12:08:44Z | ---
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
base_model: lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
tags:
- pruna-ai
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed with llm-int8.
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We use safetensors.
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check requirements from the original repo lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1 installed. In particular, check python, cuda, and transformers versions.
1. Make sure that you have installed quantization related packages.
```bash
pip install transformers accelerate bitsandbytes>0.37.0
```
2. Load & run the model.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("PrunaAI/lalainy-ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1-bnb-smashed", trust_remote_code=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained("lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1")
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
outputs = model.generate(input_ids, max_new_tokens=216)
tokenizer.decode(outputs[0])
```
## Configurations
The configuration info are in `smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model lalainy/ECE-PRYMMAL-YL-0.5B-SLERP-BIS-V1 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vijay-ravichander-Llama-1B-Code-LoRA-r32-merged-bnb-smashed | PrunaAI | 2024-11-13T13:52:09Z | 5 | 0 | null | [
"safetensors",
"llama",
"pruna-ai",
"base_model:vijay-ravichander/Llama-1B-LoRA-r32-merged",
"base_model:quantized:vijay-ravichander/Llama-1B-LoRA-r32-merged",
"8-bit",
"bitsandbytes",
"region:us"
] | null | 2024-11-12T13:11:41Z | ---
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
base_model: vijay-ravichander/Llama-1B-Code-LoRA-r32-merged
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
tags:
- pruna-ai
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed with llm-int8.
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We use safetensors.
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check requirements from the original repo vijay-ravichander/Llama-1B-Code-LoRA-r32-merged installed. In particular, check python, cuda, and transformers versions.
1. Make sure that you have installed quantization related packages.
```bash
pip install transformers accelerate bitsandbytes>0.37.0
```
2. Load & run the model.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("PrunaAI/vijay-ravichander-Llama-1B-Code-LoRA-r32-merged-bnb-smashed", trust_remote_code=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained("vijay-ravichander/Llama-1B-Code-LoRA-r32-merged")
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
outputs = model.generate(input_ids, max_new_tokens=216)
tokenizer.decode(outputs[0])
```
## Configurations
The configuration info are in `smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vijay-ravichander/Llama-1B-Code-LoRA-r32-merged before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/Pessie-menstrual-health-model-bnb-smashed | PrunaAI | 2024-11-13T13:51:35Z | 5 | 0 | null | [
"safetensors",
"llama",
"pruna-ai",
"base_model:Pessie/menstrual-health-model",
"base_model:quantized:Pessie/menstrual-health-model",
"8-bit",
"bitsandbytes",
"region:us"
] | null | 2024-11-12T13:30:34Z | ---
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
base_model: Pessie/menstrual-health-model
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
tags:
- pruna-ai
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed with llm-int8.
- ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We use safetensors.
- ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check requirements from the original repo Pessie/menstrual-health-model installed. In particular, check python, cuda, and transformers versions.
1. Make sure that you have installed quantization related packages.
```bash
pip install transformers accelerate bitsandbytes>0.37.0
```
2. Load & run the model.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("PrunaAI/Pessie-menstrual-health-model-bnb-smashed", trust_remote_code=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained("Pessie/menstrual-health-model")
input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
outputs = model.generate(input_ids, max_new_tokens=216)
tokenizer.decode(outputs[0])
```
## Configurations
The configuration info are in `smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model Pessie/menstrual-health-model before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
trollek/Holger-7B-v0.1 | trollek | 2024-11-13T13:50:07Z | 10 | 0 | transformers | [
"transformers",
"safetensors",
"llama",
"text-generation",
"mergekit",
"merge",
"conversational",
"da",
"en",
"dataset:trollek/Danoia-v01",
"dataset:trollek/ProbingPanoia-v01",
"dataset:trollek/Panoia-v02",
"dataset:smangrul/ultrachat-10k-chatml",
"dataset:Gryphe/Sonnet3.5-SlimOrcaDedupCleaned-20k",
"base_model:LumiOpen/Viking-7B",
"base_model:merge:LumiOpen/Viking-7B",
"base_model:mpasila/Viking-Magnum-v0.1-7B",
"base_model:merge:mpasila/Viking-Magnum-v0.1-7B",
"base_model:mpasila/Viking-SlimSonnet-v1-7B",
"base_model:merge:mpasila/Viking-SlimSonnet-v1-7B",
"license:apache-2.0",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-generation | 2024-11-09T01:49:53Z | ---
license: apache-2.0
datasets:
- trollek/Danoia-v01
- trollek/ProbingPanoia-v01
- trollek/Panoia-v02
- smangrul/ultrachat-10k-chatml
- Gryphe/Sonnet3.5-SlimOrcaDedupCleaned-20k
language:
- da
- en
base_model:
- LumiOpen/Viking-7B
- mpasila/Viking-Magnum-v0.1-7B
- mpasila/Viking-SlimSonnet-v1-7B
library_name: transformers
tags:
- mergekit
- merge
---
<img src="https://huggingface.co/trollek/Holger-7B-v0.1/resolve/main/holger-viking.png" alt="Android Vikingen Holger">
# Du har fundet Holger!
Det er *forrykt* at danske open source sprogmodeller halter lidt; **æøå** be damned! Og nok også andre ting, men \[A-ZÆØÅa-zæøå\] giver kronisk mentaleksem. **Anyway!**
HOLGER! En fintunet åben Viking på åbne data under et åbent licens. Respekt for licenser er sådan kapitalismen er, og GUDS NÅDE fucking trøste dig hvis du ikke udgiver Llama modeller uden også at bøje knæet for Zuck. Redemption arc eller ej. 10% af 1 fantasiliard er stadig mere end jeg kan prostiturere mig selv for.
## Sprogmodellens beskrivelse
> Hva' lav' do nu? - Det vil jeg helst ikke snakke om. - Kom nu! - Jeg træner og snakker med sprogmodeller.. - Fra udlandet? - Jaaaa...
Viking modellen kan åbenbart finde ud af at overføre træning fra engelsk til dansk, og sandsynligvis de andre skandisprog. Maybe.
## Kvanter
- [trollek/Holger-7B-v0.1-GGUF](https://huggingface.co/trollek/Holger-7B-v0.1-GGUF)
## Merge Detaljer
### Merge Metode
Denne model er merged med della_linear metoden og med [LumiOpen/Viking-7B](https://huggingface.co/LumiOpen/Viking-7B) som base.
### Modeller Merged
Følgende modeller er brugt i dette merge:
* viking/merges/holger5
* viking/merges/holger3
* [mpasila/Viking-Magnum-v0.1-7B](https://huggingface.co/mpasila/Viking-Magnum-v0.1-7B)
* [mpasila/Viking-SlimSonnet-v1-7B](https://huggingface.co/mpasila/Viking-SlimSonnet-v1-7B)
### Konfiguration
Følgende YAML konfigurationsfil blev brugt til at skabe dette merge:
```yaml
models:
- model: viking/merges/holger3
parameters:
weight: 0.53
density: 0.55
epsilon: 0.11
- model: viking/merges/holger5
parameters:
weight: 0.78
density: 0.81
epsilon: 0.17
- model: mpasila/Viking-Magnum-v0.1-7B
parameters:
weight: 0.88
density: 0.91
epsilon: 0.07
- model: mpasila/Viking-SlimSonnet-v1-7B
parameters:
weight: 0.83
density: 0.75
epsilon: 0.11
merge_method: della_linear
base_model: LumiOpen/Viking-7B
parameters:
normalize: true
int8_mask: true
lambda: 1.08
dtype: bfloat16
``` |
homeb82784/Qwen2-7B-Instruct-FinanceSFT | homeb82784 | 2024-11-13T13:45:17Z | 5 | 0 | transformers | [
"transformers",
"safetensors",
"qwen2",
"text-generation",
"text-generation-inference",
"unsloth",
"trl",
"krx",
"conversational",
"en",
"license:apache-2.0",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | text-generation | 2024-11-07T05:17:48Z | ---
base_model: unsloth/qwen2-7b-instruct-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- qwen2
- trl
- krx
license: apache-2.0
language:
- en
---
# Uploaded model
- **Developed by:** homeb82784
- **License:** apache-2.0
- **Finetuned from model :** unsloth/qwen2-7b-instruct-bnb-4bit
This qwen2 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth) |
susmitabhatt/xlsr-nm-nomo | susmitabhatt | 2024-11-13T13:42:40Z | 5 | 0 | transformers | [
"transformers",
"tensorboard",
"safetensors",
"wav2vec2",
"automatic-speech-recognition",
"generated_from_trainer",
"base_model:facebook/wav2vec2-large-xlsr-53",
"base_model:finetune:facebook/wav2vec2-large-xlsr-53",
"license:apache-2.0",
"endpoints_compatible",
"region:us"
] | automatic-speech-recognition | 2024-11-13T12:41:32Z | ---
library_name: transformers
license: apache-2.0
base_model: facebook/wav2vec2-large-xlsr-53
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: xlsr-nm-nomo
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# xlsr-nm-nomo
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0423
- Wer: 0.3916
## Model description
More information needed
## Intended uses & limitations
More information needed
## Training and evaluation data
More information needed
## Training procedure
### Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0004
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 132
- num_epochs: 100
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-------:|:----:|:---------------:|:------:|
| 4.7334 | 6.4590 | 200 | 3.1082 | 1.0 |
| 2.9544 | 12.9180 | 400 | 2.8429 | 0.9956 |
| 2.0771 | 19.3607 | 600 | 1.2204 | 0.8341 |
| 0.7271 | 25.8197 | 800 | 1.0868 | 0.5531 |
| 0.3103 | 32.2623 | 1000 | 1.0536 | 0.4912 |
| 0.1852 | 38.7213 | 1200 | 0.9030 | 0.4469 |
| 0.1399 | 45.1639 | 1400 | 0.8980 | 0.4491 |
| 0.0864 | 51.6230 | 1600 | 0.8315 | 0.4292 |
| 0.0643 | 58.0656 | 1800 | 0.9488 | 0.4004 |
| 0.0525 | 64.5246 | 2000 | 0.9354 | 0.4137 |
| 0.0455 | 70.9836 | 2200 | 0.9717 | 0.4093 |
| 0.0383 | 77.4262 | 2400 | 0.9781 | 0.4004 |
| 0.0261 | 83.8852 | 2600 | 1.1244 | 0.3938 |
| 0.0265 | 90.3279 | 2800 | 1.0439 | 0.4004 |
| 0.0197 | 96.7869 | 3000 | 1.0423 | 0.3916 |
### Framework versions
- Transformers 4.47.0.dev0
- Pytorch 2.4.0
- Datasets 3.0.1
- Tokenizers 0.20.0
|
Kouskousi/gte-large-en-v1.5_SEC_docs_ft_with_5_epochs | Kouskousi | 2024-11-13T13:42:22Z | 5 | 0 | sentence-transformers | [
"sentence-transformers",
"safetensors",
"new",
"sentence-similarity",
"feature-extraction",
"generated_from_trainer",
"dataset_size:9400",
"loss:MultipleNegativesRankingLoss",
"custom_code",
"arxiv:1908.10084",
"arxiv:1705.00652",
"base_model:Alibaba-NLP/gte-large-en-v1.5",
"base_model:finetune:Alibaba-NLP/gte-large-en-v1.5",
"model-index",
"autotrain_compatible",
"text-embeddings-inference",
"endpoints_compatible",
"region:us"
] | sentence-similarity | 2024-11-13T12:59:29Z | ---
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- generated_from_trainer
- dataset_size:9400
- loss:MultipleNegativesRankingLoss
base_model: Alibaba-NLP/gte-large-en-v1.5
widget:
- source_sentence: What was the percentage change in net investment income from 2021
to 2022?
sentences:
- "Index\nAmeriprise Financial, Inc.\nConsolidated Results of Operations\nYear Ended\
\ December 31, 2022 Compared to Year Ended December 31, 2021\nThe following table\
\ presents our consolidated results of operations:\n \nYears Ended December 31,\n\
Change\n2022\n2021\n(in millions)\nRevenues\nManagement and financial advice fees\n\
$\n9,033 \n$\n9,275 \n$\n(242)\n(3)\n%\nDistribution fees\n1,939 \n1,828 \n111\
\ \n6 \nNet investment income\n1,474 \n1,683 \n(209)\n(12)\nPremiums, policy and\
\ contract charges\n1,397 \n221 \n1,176 \nNM\nOther revenues\n491 \n382 \n109\
\ \n29 \nTotal revenues\n14,334 \n13,389 \n945 \n7 \nBanking and deposit interest\
\ expense\n76 \n12 \n64 \nNM\nTotal net revenues\n14,258 \n13,377 \n881 \n7 \n\
Expenses\n \n \n \n \nDistribution expenses\n4,935 \n5,028 \n(93)\n(2)\nInterest\
\ credited to fixed accounts\n665 \n600 \n65 \n11 \nBenefits, claims, losses and\
\ settlement expenses\n242 \n(156)\n398 \nNM\nRemeasurement (gains) losses of\
\ future policy benefit reserves\n1 \n(52)\n53 \nNM\nChange in fair value of market\
\ risk benefits\n311 \n(113)\n424 \nNM\nAmortization of deferred acquisition costs\n\
252 \n259 \n(7)\n(3)\nInterest and debt expense\n198 \n191 \n7 \n4 \nGeneral and\
\ administrative expense\n3,723 \n3,435 \n288 \n8 \nTotal expenses\n10,327 \n\
9,192 \n1,135 \n12 \nPretax income\n3,931 \n4,185 \n(254)\n(6)\nIncome tax provision\n\
782 \n768 \n14 \n2 \nNet income\n$\n3,149 \n$\n3,417 \n$\n(268)\n(8)\n%\nNM Not\
\ Meaningful - variance equal to or greater than 100%.\nOverall\nPretax income\
\ decreased $254 million, or 6%, for 2022 compared to the prior year. The following\
\ impacts were significant drivers of the year-over-year change in pretax\nincome:\n\
•\nThe prior year impact of the block transfer reinsurance transaction resulted\
\ in $524 million of pretax income for 2021 primarily reflecting the net realized\
\ gains on\ninvestments sold to the reinsurer.\n•\nA negative impact from lower\
\ average equity markets compared to the prior year. Our average WEI, which is\
\ a proxy for equity movements on AUM, decreased\n7% in 2022 compared to the prior\
\ year. The average S&P 500 index was 4% lower for 2022 compared to the prior\
\ year.\n•\nThe favorable impact of unlocking was $133 million for 2022 compared\
\ to an unfavorable impact of $113 million for the prior year.\n•\nA favorable\
\ impact from the increase in short-term interest rates compared to the prior\
\ year.\n•\nThe market impact on non-traditional long-duration products (including\
\ variable and fixed deferred annuity contracts and UL insurance contracts), net\
\ of hedges\nand the reinsurance accrual was a benefit of $483 million for 2022\
\ compared to a benefit of $464 million for the prior year.\n50"
- "Intersegment revenues for this segment reflect fees paid by our Asset Management\
\ segment for marketing support and other services provided in\nconnection with\
\ the availability of VIT Funds. Intersegment expenses for this segment include\
\ distribution expenses for services provided by our Advice & Wealth\nManagement\
\ segment, as well as expenses for investment management services provided by\
\ our Asset Management segment. All intersegment activity is eliminated\nin our\
\ consolidated results.\nProtection\nWe provide life and disability income insurance\
\ products to address the protection and risk management needs of our retail clients.\
\ New \nRiverSource\n insurance\nproducts are exclusively offered through our\
\ advisor network. Our advisors also offer insurance products of unaffiliated\
\ carriers. The primary sources of revenues for\nour protection business are premiums,\
\ fees and charges we receive to assume insurance-related risk. We earn net investment\
\ income on owned assets supporting\ninsurance reserves and on capital supporting\
\ the business. We also receive fees based on the level of the RiverSource Life\
\ companies’ separate account assets\nsupporting variable universal life investment\
\ options. The protection products earn intersegment revenues from fees paid by\
\ our Asset Management segment for\nmarketing support and other services provided\
\ in connection with the availability of VIT Funds under the variable universal\
\ life contracts. Intersegment expenses for\nthe protection products include distribution\
\ expenses for services provided by our Advice & Wealth Management segment, as\
\ well as expenses for investment\nmanagement services provided by our Asset Management\
\ segment. All intersegment activity is eliminated in our consolidated results.\n\
®\n6"
- 'Index
Ameriprise Financial, Inc.
Slide 5 - Regulatory Oversight Chart.jpg
Advice & Wealth Management Regulation
Certain of our subsidiaries are registered with the SEC as broker-dealers under
the Securities Exchange Act of 1934 (“Exchange Act”) and with certain states,
the
District of Columbia and other U.S. territories. Our broker-dealer subsidiaries
are also members of self-regulatory organizations, including Financial Industry
Regulatory Authority (“FINRA”), and are subject to the regulations of these organizations.
The SEC and FINRA have stringent rules with respect to the net capital
requirements (which includes rules around customer protection) and the marketing
and trading activities of broker-dealers. Our broker-dealer subsidiaries, as well
as
our financial advisors and other personnel, must obtain all required state and
FINRA licenses and registrations to engage in the securities business and take
certain
steps to maintain such registrations in good standing. SEC regulations also impose
notice requirements and capital
11'
- source_sentence: What are the key components of Ameriprise Financial's strategy
for attracting and retaining a diverse workforce?
sentences:
- "Index\nAmeriprise Financial, Inc.\n2022 due to a $75.8 billion increase in Advice\
\ & Wealth Management AUM driven by market appreciation and wrap account net inflows\
\ and a $52.9 billion increase in\nAsset Management AUM primarily driven by market\
\ appreciation. Total AUA increased $57.5 billion, or 26%, to $279.5 billion as\
\ of December 31, 2023 compared to\nthe prior year primarily driven by equity\
\ market appreciation, and an increase in third-party money market funds and brokered\
\ CDs. See our segment results of\noperations discussion for additional information\
\ on changes in our AUM.\nConsolidated Results of Operations\nYear Ended December\
\ 31, 2023 Compared to Year Ended December 31, 2022\nThe following table presents\
\ our consolidated results of operations:\n \nYears Ended December 31,\nChange\n\
2023\n2022\n(in millions)\nRevenues\nManagement and financial advice fees\n$\n\
8,907 \n$\n9,033 \n$\n(126)\n(1)\n%\nDistribution fees\n1,931 \n1,939 \n(8)\n\
— \nNet investment income\n3,206 \n1,474 \n1,732 \nNM\nPremiums, policy and contract\
\ charges\n1,539 \n1,397 \n142 \n10 \nOther revenues\n513 \n491 \n22 \n4 \nTotal\
\ revenues\n16,096 \n14,334 \n1,762 \n12 \nBanking and deposit interest expense\n\
561 \n76 \n485 \nNM\nTotal net revenues\n15,535 \n14,258 \n1,277 \n9 \nExpenses\n\
\ \n \n \n \nDistribution expenses\n5,078 \n4,935 \n143 \n3 \nInterest credited\
\ to fixed accounts\n654 \n665 \n(11)\n(2)\nBenefits, claims, losses and settlement\
\ expenses\n1,350 \n242 \n1,108 \nNM\nRemeasurement (gains) losses of future policy\
\ benefit reserves\n(20)\n1 \n(21)\nNM\nChange in fair value of market risk benefits\n\
798 \n311 \n487 \nNM\nAmortization of deferred acquisition costs\n246 \n252 \n\
(6)\n(2)\nInterest and debt expense\n324 \n198 \n126 \n64 \nGeneral and administrative\
\ expense\n3,871 \n3,723 \n148 \n4 \nTotal benefits and expenses\n12,301 \n10,327\
\ \n1,974 \n19 \nPretax income\n3,234 \n3,931 \n(697)\n(18)\nIncome tax provision\n\
678 \n782 \n(104)\n(13)\nNet income\n$\n2,556 \n$\n3,149 \n$\n(593)\n(19)\n%\n\
NM Not Meaningful - variance equal to or greater than 100%.\nOverall\nPretax\
\ income decreased $697 million, or 18%, for 2023 compared to the prior year.\
\ The following impacts were significant drivers of the year-over-year change\
\ in\npretax income:\n•\nThe market impact on non-traditional long-duration products,\
\ net of hedges was an expense of $608 million for 2023 compared to a benefit\
\ of $483 million for the\nprior year.\n•\nThe unfavorable impact of unlocking\
\ was $99 million for 2023 compared to a favorable impact of unlocking of $133\
\ million for the prior year.\n•\nA favorable impact from the trend in rising\
\ interest rates on the investment portfolio yield, including from investment\
\ portfolio repositioning in our insurance\nbusiness in the fourth quarter of\
\ 2022, along with higher balances in bank and certificate products.\n39"
- 'At December 31, 2023 and 2022, the parent company had $544 million and $389 million,
respectively, in cash, cash equivalents, and unencumbered liquid securities.
Liquid securities predominantly include U.S. government agency mortgage back securities.
Additional sources of liquidity include a line of credit with an affiliate up
to $727 million and an unsecured revolving committed credit facility for up to
$1.0 billion that expires in June 2026. Management’s estimate of liquidity available
to the
parent company in a volatile and uncertain economic environment as of December
31, 2023 was $1.8 billion which includes cash, cash equivalents, unencumbered
liquid securities, the line of credit with an affiliate and a portion of the committed
credit facility.
Under the terms of the committed credit facility, we can increase the availability
to $1.25 billion upon satisfaction of certain approval requirements. Available
borrowings under this facility are reduced by any outstanding letters of credit.
At December 31, 2023, we had no outstanding borrowings under this credit facility
and
had $1 million of letters of credit issued against the facility. Our credit facility
contains various administrative, reporting, legal and financial covenants. We
remain in
compliance with all such covenants at December 31, 2023.
In addition, we have access to collateralized borrowings, which may include repurchase
agreements and Federal Home Loan Bank (“FHLB”) advances, and advances
at the Federal Reserve. Our subsidiaries, RiverSource Life Insurance Company (“RiverSource
Life”), and Ameriprise Bank are members of the FHLB of Des Moines,
which provides access to collateralized borrowings. As of December 31, 2023 and
2022, we had an estimated maximum borrowing capacity of $8.6 billion and $8.0
billion, respectively, of borrowing capacity under the FHLB facilities, of which
$201 million was outstanding as of both December 31, 2023 and 2022, and is
collateralized with commercial mortgage backed securities. In addition, Ameriprise
Bank maintains access to borrowings from the Federal Reserve which are
collateralized with residential mortgage backed securities, commercial mortgage
backed securities and
60'
- 'The impact of these factors on our business may vary from country to country
and certain competitors may have competitive advantage in
certain jurisdictions.
Competitors of our Retirement & Protection Solutions segment consist of both stock
and mutual insurance companies. Competitive factors affecting the sale of
variable annuity and insurance products include distribution capabilities, price,
product features and innovation, hedging capability, investment performance,
commission structure, reinsurance availability and pricing, perceived financial
strength and financial strength ratings, claims-paying ratings, technology and
service,
advertising, brand recognition and financial strength ratings from rating agencies.
Human Capital Management
Ameriprise Financial has a strong values-driven and inclusive culture that is
the foundation of all that we do. While our individual business lines serve different
client
needs, we have a common vision and values that drive our business and how we work
with clients and each other. Our values are the following:
•
Client focused;
•
Integrity always;
•
Excellence in all we do; and
•
Respect for the individuals and for the communities in which we live and work.
To ensure our long-term success, we must attract, retain, engage and develop a
diverse, high-performing workforce. We are committed to providing an excellent
employee and advisor experience for all of our people. This includes approximately
13,800 global
8'
- source_sentence: What is the significance of Columbia Threadneedle in Ameriprise's
asset management segment?
sentences:
- 'Index
Ameriprise Financial, Inc.
Expenses
Distribution expenses decreased $93 million, or 2%, for 2022 compared to the prior
year primarily lower average equity markets and decreased transactional activity.
Interest credited to fixed accounts increased $65 million, or 11%, for 2022 compared
to the prior year primarily reflecting the following items:
•
An $23 million decrease in expense from the unhedged nonperformance credit spread
risk adjustment on IUL benefits. The favorable impact of the
nonperformance credit spread was $13 million for 2022 compared to an unfavorable
impact of $10 million for the prior year.
•
A $105 million increase in expense from other market impacts on IUL benefits,
net of hedges, which was an expense of $51 million for 2022 compared to a benefit
of $54 million for the prior year. The increase in expense was primarily due to
an increase in the IUL embedded derivative in the current year period, which
reflected higher option costs due to a higher new money rate, compared to a decrease
in the IUL embedded derivative in the prior year period, which reflected
lower option costs due to higher discount rates.
Benefits, claims, losses and settlement expenses increased $398 million for 2022
compared to the prior year primarily reflecting the following items:
•
A $1.2 billion decrease in expense associated with the reinsurance transaction
for life contingent payout annuity policies in the prior year.
•
An $806 million decrease in expense from market impacts on SVA embedded derivative,
net of hedges in place to offset those risks. This decrease was the result
of a favorable $1.0 billion change in the market impact on the SVA embedded derivative,
partially offset by an unfavorable $224 million change in the market
impact on derivatives hedging the SVA embedded derivative. The main market driver
contributing to these changes was the equity market impact on the SVA
embedded derivative net of the impact on the corresponding hedge assets, which
resulted in a benefit for 2022 compared to an expense in the prior year.
Remeasurement (gains) losses of future policy benefit reserves increased $53 million
for 2022 compared to the prior year primarily reflecting an increase in expense
on
LTC insurance as policy and claim terminations returned to more normalized levels
compared to the prior year period which benefited from COVID-19 related impacts.
Change in fair value of market risk benefits increased $424 million for 2022 compared
to the prior year primarily reflecting the following items:
•
A $769 million increase in expense from other market impacts on variable annuity
guaranteed benefits, net of hedges in place to offset those risks. This decrease
was the result of an unfavorable $865 million change in the market impact on variable
annuity guaranteed benefits reserves, partially offset by a favorable
$96 million change in the market impact on derivatives hedging the variable annuity
guaranteed benefits. The main market drivers contributing to these changes
are summarized below:
•
Equity market impact on the variable annuity guaranteed benefits liability net
of the impact on the corresponding hedge assets resulted in an expense for
2022 compared to a benefit in the prior year.
•
Interest rate and bond impact on the variable annuity guaranteed benefits liability
net of the impact on the corresponding hedge assets resulted in a benefit
for 2022 compared to an expense in the prior year.
•
Volatility impact on the variable annuity guaranteed benefits liability net of
the impact on the corresponding hedge assets resulted in a higher expense for
2022 compared to the prior year.
•
Other unhedged items, including the difference between the assumed and actual
underlying separate account investment performance, transaction costs
and various behavioral items, were a higher net expense for 2022 compared to the
prior year.
General and administrative expense increased $288 million, or 8%, 2022 the prior
year primarily reflecting the operating expenses of the acquired BMO Global Asset
Management (EMEA) business and higher integration related expenses, partially
offset by disciplined expense management and reengineering, lower performance
fee
related compensation and a favorable foreign exchange impact.
Income Taxes
Our effective tax rate was 19.9% for 2022 compared to 18.3% for the prior year.
See Note 24 to our Consolidated Financial Statements for additional discussion
on
income taxes.
52'
- "Index\nAmeriprise Financial, Inc.\nIntellectual Property\nWe rely on a combination\
\ of contractual rights and copyright, trademark, and patent registrations and\
\ trade secret laws to establish and protect our intellectual\nproperty. In the\
\ U.S. and other jurisdictions, we have established and registered, or filed applications\
\ to register, certain trademarks and service marks that we\nconsider important\
\ to the marketing of our products and services, including but not limited to\
\ the \nAmeriprise Financial, Threadneedle, RiverSource, Columbia\nThreadneedle\
\ and Columbia Threadneedle Investments \nbrands. We have in the past and will\
\ continue to establish and protect our intellectual property rights.\nEnterprise\
\ Risk Management\nEnterprise risk management and our risk management program\
\ is an important component in how we manage our business. All operating subsidiaries\
\ of Ameriprise\nmust comply with our enterprise risk management policy and framework,\
\ which: (i) establishes a structure for effective enterprise risk management,\
\ including\noversight and governance; (ii) delineates key constituent roles and\
\ responsibilities; and (iii) imposes a number of core risk management processes.\
\ The enterprise risk\nmanagement policy is designed to manage risks that may\
\ impact Ameriprise, including capital, credit, market, liquidity, operational,\
\ strategic, reputational, legal and\ncompliance, and product. The enterprise\
\ risk management policy is supported by underlying risk policies at business\
\ units that provide further detail on the business\nunit’s risk governance, appetite,\
\ and tolerance.\nRegulation\nVirtually all aspects of our business, including\
\ the activities of the parent company and our subsidiaries, are subject to various\
\ federal, state, local and foreign laws\nand regulations. These laws and regulations\
\ provide broad regulatory, administrative and enforcement powers to supervisory\
\ agencies and other bodies, including\nU.S. federal and state regulatory and\
\ law enforcement agencies, foreign government agencies or regulatory bodies and\
\ U.S. and foreign securities exchanges. The\ncosts of complying with such laws\
\ and regulations are significant and increasing, and the consequences for the\
\ failure to comply may include civil or criminal\ncharges, fines, censure, the\
\ suspension of individual employees, restrictions on or prohibitions from engaging\
\ in certain lines of business (or in certain states or\ncountries), revocation\
\ of certain registrations and reputational damage. We have made and expect to\
\ need to continue to make significant investments in our\ncompliance and supervision\
\ processes, enhancing policies, procedures and oversight to monitor our compliance\
\ with the numerous legal and regulatory requirements\napplicable to our business.\n\
We operate in a highly scrutinized regulatory environment and it remains subject\
\ to change. Regulatory developments, both in and outside of the U.S., have resulted\n\
or are expected to result in greater regulatory oversight and internal compliance\
\ obligations for firms across the financial services industry. In addition, we\
\ continue\nto see enhanced legislative and regulatory interest regarding retirement\
\ investing and fiduciary initiatives, as well as environmental, social and governance\
\ (“ESG”)\nconsideration and responsible investing; cybersecurity and resilience;\
\ the use of artificial intelligence; responsible information and data collection,\
\ storage and use;\nfinancial crime prevention; and privacy matters, and we will\
\ continue to closely review and monitor any legislative or regulatory proposals\
\ and changes. States in the\nU.S. and jurisdictions outside the U.S. continue\
\ to add new complexity to the patchwork of laws and regulations already in existence\
\ relating to privacy,\ncybersecurity, artificial intelligence and other areas\
\ and we are expecting similar new laws at the federal level and in multiple states\
\ in the U.S. The same complexity\nresulting from multiple standards exists for\
\ retirement investing where individual states and federal regulators continue\
\ to propose or enact their own rules. These\nlegal and regulatory changes have\
\ impacted and may in the future impact how we are regulated and how we operate\
\ and govern our businesses.\nThe discussion and overview set forth below provides\
\ a general framework of the primary laws and regulations impacting our businesses.\
\ Certain of our subsidiaries\nmay be subject to one or more elements of this\
\ regulatory framework depending on the nature of their business, the products\
\ and services they provide and the\ngeographic locations in which they operate.\
\ To the extent the discussion includes references to statutory and regulatory\
\ provisions, it is qualified in its entirety by\nreference to these statutory\
\ and regulatory provisions and is current only as of the date of this report.\n\
10"
- 'Index
Ameriprise Financial, Inc.
term needs. As part of our goal-based approach to financial advice, our advisors
help clients actively manage investing, saving and spending so they have a more
complete picture of their financial life.
A significant portion of revenues in this segment are fee-based and driven by
the level of client assets, which is impacted by both market movements and net
flows.
We also earn revenue and income through other sources, including the following:
•
We earn net investment income on owned assets from Ameriprise Certificate Company
and Ameriprise Bank, both wholly owned subsidiaries of Ameriprise.
•
We earn financial planning fees as well as transaction and other fees.
•
We earn distribution fees for providing non-affiliated products and intersegment
revenues for providing our affiliated products and services to our retail
clients. Intersegment expenses for this segment include investment management
services provided by our Asset Management segment. All intersegment
activity is eliminated in our consolidated results.
Our Financial Advisor Platform
With more than 10,000 advisors, we are one of the top branded advisor platforms
in the U.S. market. Advisors can choose to affiliate with us in multiple ways
as
noted below, and each option offers different levels of support and compensation.
Slide 3 - Advisor Affiliations-v2.jpg
We offer the following products and services through our Advice & Wealth Management
segment:
•
Financial planning and advice services to provide personalized financial planning
and financial solutions for which we charge fees and may receive sales
commissions for selling products that aid in our clients’ plans.
•
Discretionary and non-discretionary investment advisory accounts for which we
receive fees based on the assets held in that account, as well as related fees
or
costs associated with the underlying securities held in that account.
•
Brokerage products and services for retail and institutional clients.
•
Cash management and banking products, including brokerage sweep programs, cash
management accounts, savings accounts, credit cards, margin loans and
pledged asset lines of credit.
•
Face-amount certificates through the Ameriprise Certificate Company, a wholly
owned subsidiary.
•
Mutual fund offerings from our own Columbia funds as well as approximately 135
unaffiliated mutual fund families, representing approximately 2,150 mutual funds
on our brokerage platform for which mutual fund families and other companies generally
pay us a portion of the revenue generated from sales of those funds,
administrative fees, and fees from the ongoing management attributable to our
clients’ ownership in the fund.
•
Insurance and annuities products from both RiverSource Life companies as well
as certain third parties, and we receive a portion of the revenue generated from
the sale of unaffiliated products and certain administrative fees.
Our Segments - Asset Management
Through Columbia Threadneedle, we provide investment management, advice and products
to retail, high net worth and institutional clients on a global scale.
4'
- source_sentence: How does Ameriprise Financial's distribution team support its global
asset management business?
sentences:
- "Index\nAmeriprise Financial, Inc.\nOther revenues increased $84 million, or 58%,\
\ for 2022 compared to the prior year primarily reflecting the yield on deposit\
\ receivables arising from reinsurance\ntransactions.\nExpenses\nRemeasurement\
\ (gains) losses of future policy benefit reserves increased $46 million for 2022\
\ compared to the prior year primarily reflecting an increase in expense on\n\
LTC insurance as policy and claim terminations returned to more normalized levels\
\ compared to the prior year which benefited from COVID-19 related impacts.\n\
General and administrative expense, which excludes integration and restructuring\
\ charges and expenses attributable to CIEs, \ndecreased $39 million, or 15%,\
\ to 2022\ncompared to the prior year primarily due to a larger unfavorable change\
\ in the mark-to-market impact on share-based compensation expenses in the prior\
\ year due to\nshare price appreciation.\nFair Value Measurements\nWe report certain\
\ assets and liabilities at fair value; specifically, separate account assets,\
\ derivatives, market risk benefits, embedded derivatives and most\ninvestments\
\ and cash equivalents. Fair value assumes the exchange of assets or liabilities\
\ occurs in orderly transactions and is not the result of a forced liquidation\n\
or distressed sale. We include actual market prices, or observable inputs, in\
\ our fair value measurements to the extent available. Broker quotes are obtained\
\ when\nquotes from pricing services are not available. We validate prices obtained\
\ from third parties through a variety of means such as: price variance analysis,\
\ subsequent\nsales testing, stale price review, price comparison across pricing\
\ vendors and due diligence reviews of vendors. See Note 16 to the Consolidated\
\ Financial Statements\nfor additional information on our fair value measurements.\n\
Fair Value of Liabilities and Nonperformance Risk\nCompanies are required to measure\
\ the fair value of liabilities at the price that would be received to transfer\
\ the liability to a market participant (an exit price). Since\nthere is not a\
\ market for our obligations of our market risk benefits, fixed deferred indexed\
\ annuities, structured variable annuities, and IUL insurance, we consider the\n\
assumptions participants in a hypothetical market would make to reflect an exit\
\ price. As a result, we adjust the valuation of market risk benefits, fixed deferred\n\
indexed annuities, structured variable annuities, and IUL insurance by updating\
\ certain contractholder assumptions, adding explicit margins to provide for risk,\
\ and\nadjusting the rates used to discount expected cash flows to reflect a current\
\ market estimate of our nonperformance risk. The nonperformance risk adjustment\
\ is\nbased on observable market data adjusted to estimate the risk of our life\
\ insurance company subsidiaries not fulfilling these liabilities. Consistent\
\ with general market\nconditions, this estimate resulted in a spread over the\
\ U.S. Treasury curve as of December 31, 2023. As our estimate of this spread\
\ widens or tightens, the liability will\ndecrease or increase, respectively.\
\ If this nonperformance credit spread moves to a zero spread over the U.S. Treasury\
\ curve, the reduction to future total equity\nwould be approximately $795 million,\
\ net of the reinsurance accrual and income taxes (calculated at the statutory\
\ tax rate of 21%), based on December 31, 2023\ncredit spreads.\nLiquidity and\
\ Capital Resources\nOverview\nAs of December 31, 2023 and 2022, we had Available\
\ Capital for Capital Adequacy of $5.4 billion and $5.2 billion, respectively.\
\ Available Capital for Capital Adequacy\nbest reflects the available capital\
\ resources of our operations.\nWe maintained substantial liquidity during the\
\ year ended December 31, 2023. At December 31, 2023 and 2022, we had $7.5 billion\
\ and $7.0 billion, respectively, in\ncash and cash equivalents excluding CIEs\
\ and other restricted cash on a consolidated basis.\nAt December 31, 2023 and\
\ 2022, the parent company had $544 million and $389 million, respectively, in\
\ cash, cash equivalents, and unencumbered liquid securities.\nLiquid securities\
\ predominantly include U.S. government agency mortgage back securities. Additional\
\ sources of liquidity include a line of credit with an affiliate up\nto $727\
\ million and an unsecured revolving committed credit facility for up to $1.0\
\ billion that expires in June 2026. Management’s estimate of liquidity available\
\ to the\nparent company in a volatile and uncertain economic environment as of\
\ December 31, 2023 was $1.8 billion which includes cash, cash equivalents, unencumbered\n\
liquid securities, the line of credit with an affiliate and a portion of the committed\
\ credit facility."
- "Index\nAmeriprise Financial, Inc.\nDistribution fees decreased $4 million for\
\ 2022 compared to the prior year reflecting lower average equity markets and\
\ decreased transactional activity mostly offset\nby $264 million of higher fees\
\ on off-balance sheet brokerage cash due to an increase in average short-term\
\ interest rates.\nNet investment income increased $491 million for 2022 compared\
\ to the prior year primarily due to higher average invested assets due to increased\
\ bank deposits and\nhigher investment yields on the investment portfolio supporting\
\ the bank and certificate products.\nBanking and deposit interest expense increased\
\ $64 million for 2022 compared to the prior year primarily due to higher average\
\ crediting rates on bank cash deposits\nand certificates and increased cash deposits\
\ at the bank.\nExpenses\nDistribution expenses decreased $123 million, or 3%,\
\ for 2022 compared to the prior year reflecting market depreciation and decreased\
\ transactional activity.\nGeneral and administrative expense increased $114 million,\
\ or 8%, for 2022 compared to the prior year primarily due to higher volume related\
\ expenses and\ninvestments in business growth as well as lower staffing levels\
\ and limited travel and entertainment expenses in the prior year.\nAsset Management\n\
The following table presents managed assets by type:\nDecember 31,\nChange\nAverage\n\
Change\nDecember 31,\n2022\n2021\n2022\n2021\n(in billions)\nEquity\n$\n301.2\
\ \n$\n402.9 \n$\n(101.7)\n(25)\n%\n$\n333.1 \n$\n338.3 \n$\n(5.2)\n(2)\n%\nFixed\
\ income\n210.0 \n277.0 \n(67.0)\n(24)\n232.0 \n211.8 \n20.2 \n10 \nMoney market\n\
21.9 \n10.1 \n11.8 \nNM\n17.1 \n6.5 \n10.6 \nNM\nAlternative\n33.7 \n39.9 \n(6.2)\n\
(16)\n37.3 \n25.8 \n11.5 \n45 \nHybrid and other\n17.2 \n24.2 \n(7.0)\n(29)\n\
19.8 \n22.6 \n(2.8)\n(12)\nTotal managed assets\n$\n584.0 \n$\n754.1 \n$\n(170.1)\n\
(23)\n%\n$\n639.3 \n$\n605.0 \n$\n34.3 \n6 \n%\nNM Not Meaningful - variance equal\
\ to or greater than 100%.\n Average ending balances are calculated using an average\
\ of the prior period’s ending balance and all months in the current period.\n\
\ (1)\n(1)\n55"
- "Index\nAmeriprise Financial, Inc.\n•\nInstitutional and retail separately managed\
\ accounts for a range of clients, including pension, profit-sharing, employee\
\ savings, sovereign wealth funds and\nendowment funds, accounts of large- and\
\ medium-sized businesses and governmental clients, as well as the accounts of\
\ high net worth individuals and smaller\ninstitutional clients, including tax-exempt\
\ and not-for-profit organizations for which we receive management and performance-related\
\ fees.\n•\nOther separately managed accounts, including those offered through\
\ models that represent assets under advisement.\n•\nManagement of owned assets\
\ such as assets held in the general account of our RiverSource Life companies,\
\ Ameriprise Certificate Company and Ameriprise\nBank.\n•\nManagement of CLOs,\
\ which includes providing collateral management services to special purpose vehicles\
\ that primarily invest in syndicated bank loans and\nissue multiple tranches\
\ of securities collateralized by the assets for which we earn fees based on the\
\ value of assets and performance-based fees.\n•\nPrivate funds of various types\
\ where we provide investment management and related services to private, pooled\
\ investment vehicles organized as limited\npartnerships, limited liability companies,\
\ or other entities for which we may receive fees based on the value of the assets\
\ or performance-based fees.\n•\nCollective funds and separately managed accounts\
\ sponsored by Ameriprise Trust Company (“ATC”) and offered to certain qualified\
\ institutional clients such\nas retirement, pension, and profit-sharing plans\
\ for which we receive management fees.\n•\nSub-advised accounts for certain U.S.\
\ and non-U.S. funds, private banking individually managed accounts, common trust\
\ funds, and other portfolios sponsored\nor advised by other firms for which we\
\ earn management fees and possibly performance-based fees.\nDistribution\nWe\
\ maintain distribution teams and capabilities that aid the sales, marketing,\
\ and support of the products and services of our global asset management business.\n\
These distribution activities are generally organized into two major categories:\
\ retail distribution and institutional/high net worth distribution. However,\
\ alternatives\nand certain other areas have a level of specialized distribution.\n\
Our Segments - Retirement & Protection Solutions\nRiverSource solutions are one\
\ way we deliver on the Ameriprise client experience and \nConfident Retirement\n\
\ approach. We offer clients annuities, life insurance and\ndisability income\
\ insurance products to meet their needs or current stage in life—whether that\
\ is covering essentials, ensuring lifestyle, preparing for the unexpected\nor\
\ leaving a legacy. RiverSource seeks to partner with our advisors to address\
\ clients’ goals and long-term needs at a differentiated level and provide a strong\
\ risk\nprofile.\nRetirement Solutions\nThrough our advisors, we provide \nRiverSource\n\
\ annuity products to clients to help individuals address their asset accumulation\
\ and income goals. Our advisor\nnetwork is the only distributor of new \nRiverSource\n\
\ annuity products, although advisors offer fixed, variable, and structured annuities\
\ from selected unaffiliated\ninsurers. As part of the continued evolution of\
\ the business model for our Retirement & Protection Solutions segment, we focus\
\ on the accumulation solutions\nclients want (such as the structured variable\
\ annuity, a registered index-linked annuity).\nRevenues for our variable annuity\
\ products are primarily earned as fees based on a contractholder’s benefit base,\
\ contract value or separate account values, which is\nimpacted by both market\
\ movements and net asset flows. We also earn net investment income on general\
\ account assets supporting reserves for non-life contingent\npayout annuities,\
\ structured variable annuities, certain guaranteed benefits and fixed investment\
\ options offered with variable annuities, and on capital supporting\nthe business.\
\ In addition, we receive fees charged on assets allocated to our separate accounts\
\ to cover administrative costs and a portion of the management fees\nfrom the\
\ underlying investment accounts in which assets are invested. Revenues for our\
\ payout annuities with a life contingent feature are earned as premium\nrevenue.\
\ Intersegment revenues for this segment reflect fees paid by our Asset Management\
\ segment for marketing support and other services provided in\nconnection with\
\ the availability of VIT Funds. Intersegment expenses for this segment include\
\ distribution expenses for services provided by our Advice & Wealth\nManagement\
\ segment, as well as expenses for investment management services provided by\
\ our Asset Management segment. All intersegment activity is eliminated\nin our\
\ consolidated results.\nProtection\nWe provide life and disability income insurance\
\ products to address the protection and risk management needs of our retail clients.\
\ New \nRiverSource\n insurance\nproducts are exclusively offered through our\
\ advisor network. Our advisors also offer insurance products of unaffiliated\
\ carriers. The primary sources of revenues for\nour protection business are premiums,\
\ fees and charges we receive to assume insurance-related risk. We earn net investment\
\ income on owned assets supporting\ninsurance reserves and on capital supporting\
\ the business."
- source_sentence: What critical accounting estimate is highlighted regarding the
valuation of investments in Ameriprise Financial, Inc.'s financial statements?
sentences:
- "Additionally, users can sign up to receive automatic\nnotifications when new\
\ materials are posted. The information found on the website is not incorporated\
\ by reference into this report or in any other report or\ndocument we furnish\
\ or file with the SEC.\nItem 1A. \nRisk Factors\nOur operations and financial\
\ results are subject to various risks and uncertainties, including those described\
\ below, that could have a material adverse effect on our\nbusiness, financial\
\ condition or results of operations and could cause the trading price of our\
\ common stock to decline. We believe that the following information\nidentifies\
\ the material factors affecting our company based on the information we currently\
\ know. However, the risks and uncertainties our company faces are not\nlimited\
\ to those described below. Additional risks and uncertainties not presently known\
\ to us or that we currently believe to be immaterial may also adversely affect\n\
our business.\nMarket Risks\nOur results of operations and financial condition\
\ may be adversely affected by market fluctuations and by economic, political\
\ and other factors.\nOur results of operations and financial condition may be\
\ materially affected by market fluctuations and by economic and other factors.\
\ Such factors, which can be\nglobal, regional, national or local in nature, include:\
\ (i) the level and volatility of the markets, including equity prices, interest\
\ rates, commodity prices, currency values\nand other market indices and drivers;\
\ \n(ii) geopolitical strain, terrorism and armed conflicts, (iii) political dynamics\
\ or elections and social, economic and market\nconditions; (iv) the availability\
\ and cost of capital; (v) global health emergencies (such as the coronavirus\
\ disease 2019 (“COVID-19”) pandemic); (vi) technological\nchanges and\n16"
- "Index\nAmeriprise Financial, Inc.\nThe following table reconciles net income\
\ to adjusted operating earnings and the five-point average of quarter-end equity\
\ to adjusted operating equity:\n \nYears Ended December 31,\n2023\n2022\n(in\
\ millions)\nNet income\n$\n2,556 \n$\n3,149 \nLess: Adjustments \n(555)\n264\
\ \nAdjusted operating earnings\n$\n3,111 \n$\n2,885 \nTotal Ameriprise Financial,\
\ Inc. shareholders’ equity\n$\n4,116 \n$\n4,170 \nLess: AOCI, net of tax\n(2,297)\n\
(1,769)\nTotal Ameriprise Financial, Inc. shareholders’ equity, excluding AOCI\n\
6,413 \n5,939 \nLess: Equity impacts attributable to CIEs\n(4)\n— \nAdjusted operating\
\ equity\n$\n6,417 \n$\n5,939 \nReturn on equity, excluding AOCI\n39.9 \n%\n53.0\
\ \n%\nAdjusted operating return on equity, excluding AOCI\n48.5 \n%\n48.6 \n\
%\nAdjustments reflect the sum of after-tax net realized investment gains/losses,\
\ net of the reinsurance accrual; the market impact on non-traditional long-duration\
\ products (including\nvariable and fixed deferred annuity contracts and UL insurance\
\ contracts), net of hedges and the reinsurance accrual; mean reversion related\
\ impacts; block transfer reinsurance\ntransaction impacts; the market impact\
\ of hedges to offset interest rate and currency changes on unrealized gains or\
\ losses for certain investments; gain or loss on disposal of a\nbusiness that\
\ is not considered discontinued operations; integration and restructuring charges;\
\ income (loss) from discontinued operations; and net income (loss) from consolidated\n\
investment entities. After-tax is calculated using the statutory tax rate of 21%.\n\
Adjusted operating return on equity, excluding AOCI is calculated using adjusted\
\ operating earnings in the numerator and Ameriprise Financial shareholders’ equity,\
\ excluding AOCI\nand the impact of consolidating investment entities using a\
\ five-point average of quarter-end equity in the denominator. After-tax is calculated\
\ using the statutory rate of 21%.\nThe following table reconciles GAAP total\
\ equity to Available Capital for Capital Adequacy:\nDecember 31, 2023\nDecember\
\ 31, 2022\n(in millions)\nAmeriprise Financial, Inc. GAAP total equity\n$\n4,729\
\ \n$\n3,803 \nLess: AOCI\n(1,766)\n(2,546)\nAmeriprise Financial, Inc. GAAP total\
\ equity, excluding AOCI\n6,495 \n6,349 \nLess: RiverSource Life Insurance Company\
\ GAAP equity, excluding AOCI\n1,851 \n2,057 \nAdd: RiverSource Life Insurance\
\ Company statutory total adjusted capital\n3,093 \n3,103 \nLess: Goodwill and\
\ intangibles\n2,622 \n2,485 \nAdd: Other adjustments\n303 \n299 \n Available\
\ Capital for Capital Adequacy\n$\n5,418 \n$\n5,209 \nCritical Accounting Estimates\n\
The accounting and reporting policies that we use affect our Consolidated Financial\
\ Statements. Certain of our accounting and reporting policies are critical to\
\ an\nunderstanding of our consolidated results of operations and financial condition\
\ and, in some cases, the application of these policies can be significantly affected\
\ by\nthe estimates, judgments and assumptions made by management during the preparation\
\ of our Consolidated Financial Statements. The accounting and reporting\npolicies\
\ and estimates we have identified as fundamental to a full understanding of our\
\ consolidated results of operations and financial condition are described\nbelow.\
\ See Note 2 to our Consolidated Financial Statements for further information\
\ about our accounting policies.\nValuation of Investments\nThe most significant\
\ component of our investments is our Available-for-Sale securities, which we\
\ carry at fair value within our Consolidated Balance Sheets. See\nNote 16 to\
\ our Consolidated Financial Statements for discussion of the fair value of our\
\ Available-for-Sale securities. Financial markets are subject to significant\n\
movements in valuation and liquidity, which can impact our ability to liquidate\
\ and the selling price that can be realized for our securities and increases\
\ the use of\njudgment in determining the estimated fair value of certain investments.\
\ We are unable to predict impacts and determine sensitivities in reported amounts\
\ reflecting\nsuch market movements on our aggregate Available-for-Sale portfolio.\
\ Changes to these assumptions do not occur in isolation and it is impracticable\
\ to predict such\nimpacts at the individual security unit of measure which are\
\ predominately Level 2 fair value and based on observable inputs.\n(1)\n (2)\n\
(1) \n(2) \n34"
- 'Our comprehensive hedging program focuses mainly
on first order sensitivities of assets and liabilities: Equity Market Level (Delta),
Interest Rate Level (Rho) and Volatility (Vega). Additionally, various second
order
sensitivities are managed. We use various options, swaptions, swaps and futures
to manage risk exposures. The exposures are measured and monitored daily, and
adjustments to the hedge portfolio are made as necessary.
To evaluate interest rate and equity price risk we perform sensitivity testing
which measures the impact on pretax income from the sources listed below for a
12-month
period following a hypothetical 100 basis point increase in interest rates or
a hypothetical 10% decline in equity prices. The interest rate risk test assumes
a sudden
100 basis point parallel shift in the yield curve, with rates then staying at
those levels for the next 12 months. The equity price risk test assumes a sudden
10% drop in
equity prices, with equity prices then staying at those levels for the next 12
months. In estimating the values of variable annuities, indexed annuities, stock
market
certificates, indexed universal life (“IUL”) insurance and the associated hedging
instruments, we assume no change in implied market volatility despite the 10%
drop
in equity prices.
65'
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
- cosine_accuracy@1
- cosine_accuracy@3
- cosine_accuracy@5
- cosine_accuracy@10
- cosine_precision@1
- cosine_precision@3
- cosine_precision@5
- cosine_precision@10
- cosine_recall@1
- cosine_recall@3
- cosine_recall@5
- cosine_recall@10
- cosine_ndcg@10
- cosine_mrr@10
- cosine_map@100
- dot_accuracy@1
- dot_accuracy@3
- dot_accuracy@5
- dot_accuracy@10
- dot_precision@1
- dot_precision@3
- dot_precision@5
- dot_precision@10
- dot_recall@1
- dot_recall@3
- dot_recall@5
- dot_recall@10
- dot_ndcg@10
- dot_mrr@10
- dot_map@100
model-index:
- name: SentenceTransformer based on Alibaba-NLP/gte-large-en-v1.5
results:
- task:
type: information-retrieval
name: Information Retrieval
dataset:
name: Unknown
type: unknown
metrics:
- type: cosine_accuracy@1
value: 0.6062146892655367
name: Cosine Accuracy@1
- type: cosine_accuracy@3
value: 0.8508474576271187
name: Cosine Accuracy@3
- type: cosine_accuracy@5
value: 0.9225988700564972
name: Cosine Accuracy@5
- type: cosine_accuracy@10
value: 0.9711864406779661
name: Cosine Accuracy@10
- type: cosine_precision@1
value: 0.6062146892655367
name: Cosine Precision@1
- type: cosine_precision@3
value: 0.28361581920903955
name: Cosine Precision@3
- type: cosine_precision@5
value: 0.18451977401129946
name: Cosine Precision@5
- type: cosine_precision@10
value: 0.09711864406779663
name: Cosine Precision@10
- type: cosine_recall@1
value: 0.6062146892655367
name: Cosine Recall@1
- type: cosine_recall@3
value: 0.8508474576271187
name: Cosine Recall@3
- type: cosine_recall@5
value: 0.9225988700564972
name: Cosine Recall@5
- type: cosine_recall@10
value: 0.9711864406779661
name: Cosine Recall@10
- type: cosine_ndcg@10
value: 0.7958373595232834
name: Cosine Ndcg@10
- type: cosine_mrr@10
value: 0.7385916061339796
name: Cosine Mrr@10
- type: cosine_map@100
value: 0.7402573835068733
name: Cosine Map@100
- type: dot_accuracy@1
value: 0.6062146892655367
name: Dot Accuracy@1
- type: dot_accuracy@3
value: 0.8502824858757062
name: Dot Accuracy@3
- type: dot_accuracy@5
value: 0.9225988700564972
name: Dot Accuracy@5
- type: dot_accuracy@10
value: 0.9711864406779661
name: Dot Accuracy@10
- type: dot_precision@1
value: 0.6062146892655367
name: Dot Precision@1
- type: dot_precision@3
value: 0.2834274952919021
name: Dot Precision@3
- type: dot_precision@5
value: 0.18451977401129946
name: Dot Precision@5
- type: dot_precision@10
value: 0.09711864406779662
name: Dot Precision@10
- type: dot_recall@1
value: 0.6062146892655367
name: Dot Recall@1
- type: dot_recall@3
value: 0.8502824858757062
name: Dot Recall@3
- type: dot_recall@5
value: 0.9225988700564972
name: Dot Recall@5
- type: dot_recall@10
value: 0.9711864406779661
name: Dot Recall@10
- type: dot_ndcg@10
value: 0.7952805938024372
name: Dot Ndcg@10
- type: dot_mrr@10
value: 0.7378952560308499
name: Dot Mrr@10
- type: dot_map@100
value: 0.7395749514077584
name: Dot Map@100
---
# SentenceTransformer based on Alibaba-NLP/gte-large-en-v1.5
This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Alibaba-NLP/gte-large-en-v1.5](https://huggingface.co/Alibaba-NLP/gte-large-en-v1.5). It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
## Model Details
### Model Description
- **Model Type:** Sentence Transformer
- **Base model:** [Alibaba-NLP/gte-large-en-v1.5](https://huggingface.co/Alibaba-NLP/gte-large-en-v1.5) <!-- at revision 104333d6af6f97649377c2afbde10a7704870c7b -->
- **Maximum Sequence Length:** 8192 tokens
- **Output Dimensionality:** 1024 tokens
- **Similarity Function:** Cosine Similarity
<!-- - **Training Dataset:** Unknown -->
<!-- - **Language:** Unknown -->
<!-- - **License:** Unknown -->
### Model Sources
- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
### Full Model Architecture
```
SentenceTransformer(
(0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NewModel
(1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)
```
## Usage
### Direct Usage (Sentence Transformers)
First install the Sentence Transformers library:
```bash
pip install -U sentence-transformers
```
Then you can load this model and run inference.
```python
from sentence_transformers import SentenceTransformer
# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
"What critical accounting estimate is highlighted regarding the valuation of investments in Ameriprise Financial, Inc.'s financial statements?",
'Index\nAmeriprise Financial, Inc.\nThe following table reconciles net income to adjusted operating earnings and the five-point average of quarter-end equity to adjusted operating equity:\n \nYears Ended December 31,\n2023\n2022\n(in millions)\nNet income\n$\n2,556 \n$\n3,149 \nLess: Adjustments \n(555)\n264 \nAdjusted operating earnings\n$\n3,111 \n$\n2,885 \nTotal Ameriprise Financial, Inc. shareholders’ equity\n$\n4,116 \n$\n4,170 \nLess: AOCI, net of tax\n(2,297)\n(1,769)\nTotal Ameriprise Financial, Inc. shareholders’ equity, excluding AOCI\n6,413 \n5,939 \nLess: Equity impacts attributable to CIEs\n(4)\n— \nAdjusted operating equity\n$\n6,417 \n$\n5,939 \nReturn on equity, excluding AOCI\n39.9 \n%\n53.0 \n%\nAdjusted operating return on equity, excluding AOCI\n48.5 \n%\n48.6 \n%\nAdjustments reflect the sum of after-tax net realized investment gains/losses, net of the reinsurance accrual; the market impact on non-traditional long-duration products (including\nvariable and fixed deferred annuity contracts and UL insurance contracts), net of hedges and the reinsurance accrual; mean reversion related impacts; block transfer reinsurance\ntransaction impacts; the market impact of hedges to offset interest rate and currency changes on unrealized gains or losses for certain investments; gain or loss on disposal of a\nbusiness that is not considered discontinued operations; integration and restructuring charges; income (loss) from discontinued operations; and net income (loss) from consolidated\ninvestment entities. After-tax is calculated using the statutory tax rate of 21%.\nAdjusted operating return on equity, excluding AOCI is calculated using adjusted operating earnings in the numerator and Ameriprise Financial shareholders’ equity, excluding AOCI\nand the impact of consolidating investment entities using a five-point average of quarter-end equity in the denominator. After-tax is calculated using the statutory rate of 21%.\nThe following table reconciles GAAP total equity to Available Capital for Capital Adequacy:\nDecember 31, 2023\nDecember 31, 2022\n(in millions)\nAmeriprise Financial, Inc. GAAP total equity\n$\n4,729 \n$\n3,803 \nLess: AOCI\n(1,766)\n(2,546)\nAmeriprise Financial, Inc. GAAP total equity, excluding AOCI\n6,495 \n6,349 \nLess: RiverSource Life Insurance Company GAAP equity, excluding AOCI\n1,851 \n2,057 \nAdd: RiverSource Life Insurance Company statutory total adjusted capital\n3,093 \n3,103 \nLess: Goodwill and intangibles\n2,622 \n2,485 \nAdd: Other adjustments\n303 \n299 \n Available Capital for Capital Adequacy\n$\n5,418 \n$\n5,209 \nCritical Accounting Estimates\nThe accounting and reporting policies that we use affect our Consolidated Financial Statements. Certain of our accounting and reporting policies are critical to an\nunderstanding of our consolidated results of operations and financial condition and, in some cases, the application of these policies can be significantly affected by\nthe estimates, judgments and assumptions made by management during the preparation of our Consolidated Financial Statements. The accounting and reporting\npolicies and estimates we have identified as fundamental to a full understanding of our consolidated results of operations and financial condition are described\nbelow. See Note 2 to our Consolidated Financial Statements for further information about our accounting policies.\nValuation of Investments\nThe most significant component of our investments is our Available-for-Sale securities, which we carry at fair value within our Consolidated Balance Sheets. See\nNote 16 to our Consolidated Financial Statements for discussion of the fair value of our Available-for-Sale securities. Financial markets are subject to significant\nmovements in valuation and liquidity, which can impact our ability to liquidate and the selling price that can be realized for our securities and increases the use of\njudgment in determining the estimated fair value of certain investments. We are unable to predict impacts and determine sensitivities in reported amounts reflecting\nsuch market movements on our aggregate Available-for-Sale portfolio. Changes to these assumptions do not occur in isolation and it is impracticable to predict such\nimpacts at the individual security unit of measure which are predominately Level 2 fair value and based on observable inputs.\n(1)\n (2)\n(1) \n(2) \n34',
'Additionally, users can sign up to receive automatic\nnotifications when new materials are posted. The information found on the website is not incorporated by reference into this report or in any other report or\ndocument we furnish or file with the SEC.\nItem 1A. \nRisk Factors\nOur operations and financial results are subject to various risks and uncertainties, including those described below, that could have a material adverse effect on our\nbusiness, financial condition or results of operations and could cause the trading price of our common stock to decline. We believe that the following information\nidentifies the material factors affecting our company based on the information we currently know. However, the risks and uncertainties our company faces are not\nlimited to those described below. Additional risks and uncertainties not presently known to us or that we currently believe to be immaterial may also adversely affect\nour business.\nMarket Risks\nOur results of operations and financial condition may be adversely affected by market fluctuations and by economic, political and other factors.\nOur results of operations and financial condition may be materially affected by market fluctuations and by economic and other factors. Such factors, which can be\nglobal, regional, national or local in nature, include: (i) the level and volatility of the markets, including equity prices, interest rates, commodity prices, currency values\nand other market indices and drivers; \n(ii) geopolitical strain, terrorism and armed conflicts, (iii) political dynamics or elections and social, economic and market\nconditions; (iv) the availability and cost of capital; (v) global health emergencies (such as the coronavirus disease 2019 (“COVID-19”) pandemic); (vi) technological\nchanges and\n16',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]
# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]
```
<!--
### Direct Usage (Transformers)
<details><summary>Click to see the direct usage in Transformers</summary>
</details>
-->
<!--
### Downstream Usage (Sentence Transformers)
You can finetune this model on your own dataset.
<details><summary>Click to expand</summary>
</details>
-->
<!--
### Out-of-Scope Use
*List how the model may foreseeably be misused and address what users ought not to do with the model.*
-->
## Evaluation
### Metrics
#### Information Retrieval
* Evaluated with [<code>InformationRetrievalEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.InformationRetrievalEvaluator)
| Metric | Value |
|:--------------------|:-----------|
| cosine_accuracy@1 | 0.6062 |
| cosine_accuracy@3 | 0.8508 |
| cosine_accuracy@5 | 0.9226 |
| cosine_accuracy@10 | 0.9712 |
| cosine_precision@1 | 0.6062 |
| cosine_precision@3 | 0.2836 |
| cosine_precision@5 | 0.1845 |
| cosine_precision@10 | 0.0971 |
| cosine_recall@1 | 0.6062 |
| cosine_recall@3 | 0.8508 |
| cosine_recall@5 | 0.9226 |
| cosine_recall@10 | 0.9712 |
| cosine_ndcg@10 | 0.7958 |
| cosine_mrr@10 | 0.7386 |
| **cosine_map@100** | **0.7403** |
| dot_accuracy@1 | 0.6062 |
| dot_accuracy@3 | 0.8503 |
| dot_accuracy@5 | 0.9226 |
| dot_accuracy@10 | 0.9712 |
| dot_precision@1 | 0.6062 |
| dot_precision@3 | 0.2834 |
| dot_precision@5 | 0.1845 |
| dot_precision@10 | 0.0971 |
| dot_recall@1 | 0.6062 |
| dot_recall@3 | 0.8503 |
| dot_recall@5 | 0.9226 |
| dot_recall@10 | 0.9712 |
| dot_ndcg@10 | 0.7953 |
| dot_mrr@10 | 0.7379 |
| dot_map@100 | 0.7396 |
<!--
## Bias, Risks and Limitations
*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
-->
<!--
### Recommendations
*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
-->
## Training Details
### Training Dataset
#### Unnamed Dataset
* Size: 9,400 training samples
* Columns: <code>sentence_0</code> and <code>sentence_1</code>
* Approximate statistics based on the first 1000 samples:
| | sentence_0 | sentence_1 |
|:--------|:-----------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------|
| type | string | string |
| details | <ul><li>min: 12 tokens</li><li>mean: 22.99 tokens</li><li>max: 44 tokens</li></ul> | <ul><li>min: 107 tokens</li><li>mean: 671.18 tokens</li><li>max: 977 tokens</li></ul> |
* Samples:
| sentence_0 | sentence_1 |
|:--------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| <code>What is the fiscal year end date for Ameriprise Financial, Inc. as stated in the document?</code> | <code>UNITED STATES<br>SECURITIES AND EXCHANGE COMMISSION<br>WASHINGTON, D.C. 20549<br>FORM<br>10-K<br>ANNUAL REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE ACT<br>OF 1934<br>For the Fiscal Year Ended<br>December 31<br>, 2023<br>OR<br>TRANSITION REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE<br>ACT OF 1934<br>For the Transition Period from_______________________to_______________________<br>Commission File No.<br>1-32525<br>AMERIPRISE FINANCIAL, INC.<br>(Exact name of registrant as specified in its charter)<br>Delaware<br> <br>13-3180631<br>(State or other jurisdiction of incorporation or organization)<br>(I.R.S. Employer Identification No.)<br>1099 Ameriprise Financial Center<br>Minneapolis<br>Minnesota<br>55474<br> (Address of principal executive offices)<br>(Zip Code)<br>Registrant’s telephone number, including area code:<br>(612)<br>671-3131<br>Securities registered pursuant to Section 12(b) of the Act:<br>Title of each class<br>Trading Symbol<br> <br>Name of each exchange on which registered<br>Common Stock (par value $.01 per share)<br>AMP<br>New York Stock Exchange<br>Securities registered pursuant to Section 12(g) of the Act: None<br>Indicate by check mark if the registrant is a well-known seasoned issuer, as defined in Rule 405 of the Securities Act.<br>Yes<br>No<br>Indicate by check mark if the registrant is not required to file reports pursuant to Section 13 or Section 15(d) of the Exchange Act.<br>Yes<br>No<br>Indicate by check mark whether the registrant (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities Exchange<br>Act of 1934 during the preceding 12 months (or for such shorter period that the registrant was required to file such reports), and (2) has<br>been subject to such filing requirements for the past 90 days. <br>Yes<br>No<br>Indicate by check mark whether the registrant has submitted electronically every Interactive Data File required to be submitted pursuant to<br>Rule 405 of Regulation S-T (§232.405 of this chapter) during the preceding 12 months (or for such shorter period that the registrant was<br>required to submit such files). <br>Yes<br>No<br>Indicate by check mark whether the registrant is a large accelerated filer, an accelerated filer, a non-accelerated filer, smaller reporting company, or an emerging<br>growth company. See the definitions of “large accelerated filer,” “accelerated filer,” “smaller reporting company,” and “emerging growth company” in Rule 12b-2 of<br>the Exchange Act.<br>Large Accelerated Filer<br>Accelerated Filer<br>Non-accelerated Filer<br>Smaller reporting company<br>Emerging growth company<br>If an emerging growth company, indicate by check mark if the registrant has elected not to use the extended transition period for complying with any new or<br>revised financial accounting standards provided pursuant to Section 13(a) of the Exchange Act.<br>Indicate by check mark whether the registrant has filed a report on and attestation to its management’s assessment of the effectiveness of its internal control over financial<br>reporting under Section 404(b) of the Sarbanes-Oxley Act by the registered public accounting firm that prepared or issued its audit report.<br>If securities are registered pursuant to Section 12(b) of the Act, indicate by check mark whether the financial statements of the registrant included in the filing<br>reflect the correction of an error to previously issued financial statements.<br>Indicate by check mark whether any of those error corrections are restatements that required a recovery analysis of incentive-based compensation received by<br>any of the registrant’s executive officers during the relevant recovery period pursuant to §240.10D-1(b).<br>Indicate by check mark whether the registrant is a shell company (as defined in Rule 12b-2 of the Exchange Act). <br>Yes<br>No<br>The aggregate market value, as of June 30, 2023, of voting shares held by non-affiliates of the registrant was approximately $<br>34.1<br> billion.</code> |
| <code>What is the Commission File Number for Ameriprise Financial, Inc.?</code> | <code>UNITED STATES<br>SECURITIES AND EXCHANGE COMMISSION<br>WASHINGTON, D.C. 20549<br>FORM<br>10-K<br>ANNUAL REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE ACT<br>OF 1934<br>For the Fiscal Year Ended<br>December 31<br>, 2023<br>OR<br>TRANSITION REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE<br>ACT OF 1934<br>For the Transition Period from_______________________to_______________________<br>Commission File No.<br>1-32525<br>AMERIPRISE FINANCIAL, INC.<br>(Exact name of registrant as specified in its charter)<br>Delaware<br> <br>13-3180631<br>(State or other jurisdiction of incorporation or organization)<br>(I.R.S. Employer Identification No.)<br>1099 Ameriprise Financial Center<br>Minneapolis<br>Minnesota<br>55474<br> (Address of principal executive offices)<br>(Zip Code)<br>Registrant’s telephone number, including area code:<br>(612)<br>671-3131<br>Securities registered pursuant to Section 12(b) of the Act:<br>Title of each class<br>Trading Symbol<br> <br>Name of each exchange on which registered<br>Common Stock (par value $.01 per share)<br>AMP<br>New York Stock Exchange<br>Securities registered pursuant to Section 12(g) of the Act: None<br>Indicate by check mark if the registrant is a well-known seasoned issuer, as defined in Rule 405 of the Securities Act.<br>Yes<br>No<br>Indicate by check mark if the registrant is not required to file reports pursuant to Section 13 or Section 15(d) of the Exchange Act.<br>Yes<br>No<br>Indicate by check mark whether the registrant (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities Exchange<br>Act of 1934 during the preceding 12 months (or for such shorter period that the registrant was required to file such reports), and (2) has<br>been subject to such filing requirements for the past 90 days. <br>Yes<br>No<br>Indicate by check mark whether the registrant has submitted electronically every Interactive Data File required to be submitted pursuant to<br>Rule 405 of Regulation S-T (§232.405 of this chapter) during the preceding 12 months (or for such shorter period that the registrant was<br>required to submit such files). <br>Yes<br>No<br>Indicate by check mark whether the registrant is a large accelerated filer, an accelerated filer, a non-accelerated filer, smaller reporting company, or an emerging<br>growth company. See the definitions of “large accelerated filer,” “accelerated filer,” “smaller reporting company,” and “emerging growth company” in Rule 12b-2 of<br>the Exchange Act.<br>Large Accelerated Filer<br>Accelerated Filer<br>Non-accelerated Filer<br>Smaller reporting company<br>Emerging growth company<br>If an emerging growth company, indicate by check mark if the registrant has elected not to use the extended transition period for complying with any new or<br>revised financial accounting standards provided pursuant to Section 13(a) of the Exchange Act.<br>Indicate by check mark whether the registrant has filed a report on and attestation to its management’s assessment of the effectiveness of its internal control over financial<br>reporting under Section 404(b) of the Sarbanes-Oxley Act by the registered public accounting firm that prepared or issued its audit report.<br>If securities are registered pursuant to Section 12(b) of the Act, indicate by check mark whether the financial statements of the registrant included in the filing<br>reflect the correction of an error to previously issued financial statements.<br>Indicate by check mark whether any of those error corrections are restatements that required a recovery analysis of incentive-based compensation received by<br>any of the registrant’s executive officers during the relevant recovery period pursuant to §240.10D-1(b).<br>Indicate by check mark whether the registrant is a shell company (as defined in Rule 12b-2 of the Exchange Act). <br>Yes<br>No<br>The aggregate market value, as of June 30, 2023, of voting shares held by non-affiliates of the registrant was approximately $<br>34.1<br> billion.</code> |
| <code>What is the trading symbol for Ameriprise Financial, Inc. on the New York Stock Exchange?</code> | <code>UNITED STATES<br>SECURITIES AND EXCHANGE COMMISSION<br>WASHINGTON, D.C. 20549<br>FORM<br>10-K<br>ANNUAL REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE ACT<br>OF 1934<br>For the Fiscal Year Ended<br>December 31<br>, 2023<br>OR<br>TRANSITION REPORT PURSUANT TO SECTION 13 OR 15(d) OF THE SECURITIES EXCHANGE<br>ACT OF 1934<br>For the Transition Period from_______________________to_______________________<br>Commission File No.<br>1-32525<br>AMERIPRISE FINANCIAL, INC.<br>(Exact name of registrant as specified in its charter)<br>Delaware<br> <br>13-3180631<br>(State or other jurisdiction of incorporation or organization)<br>(I.R.S. Employer Identification No.)<br>1099 Ameriprise Financial Center<br>Minneapolis<br>Minnesota<br>55474<br> (Address of principal executive offices)<br>(Zip Code)<br>Registrant’s telephone number, including area code:<br>(612)<br>671-3131<br>Securities registered pursuant to Section 12(b) of the Act:<br>Title of each class<br>Trading Symbol<br> <br>Name of each exchange on which registered<br>Common Stock (par value $.01 per share)<br>AMP<br>New York Stock Exchange<br>Securities registered pursuant to Section 12(g) of the Act: None<br>Indicate by check mark if the registrant is a well-known seasoned issuer, as defined in Rule 405 of the Securities Act.<br>Yes<br>No<br>Indicate by check mark if the registrant is not required to file reports pursuant to Section 13 or Section 15(d) of the Exchange Act.<br>Yes<br>No<br>Indicate by check mark whether the registrant (1) has filed all reports required to be filed by Section 13 or 15(d) of the Securities Exchange<br>Act of 1934 during the preceding 12 months (or for such shorter period that the registrant was required to file such reports), and (2) has<br>been subject to such filing requirements for the past 90 days. <br>Yes<br>No<br>Indicate by check mark whether the registrant has submitted electronically every Interactive Data File required to be submitted pursuant to<br>Rule 405 of Regulation S-T (§232.405 of this chapter) during the preceding 12 months (or for such shorter period that the registrant was<br>required to submit such files). <br>Yes<br>No<br>Indicate by check mark whether the registrant is a large accelerated filer, an accelerated filer, a non-accelerated filer, smaller reporting company, or an emerging<br>growth company. See the definitions of “large accelerated filer,” “accelerated filer,” “smaller reporting company,” and “emerging growth company” in Rule 12b-2 of<br>the Exchange Act.<br>Large Accelerated Filer<br>Accelerated Filer<br>Non-accelerated Filer<br>Smaller reporting company<br>Emerging growth company<br>If an emerging growth company, indicate by check mark if the registrant has elected not to use the extended transition period for complying with any new or<br>revised financial accounting standards provided pursuant to Section 13(a) of the Exchange Act.<br>Indicate by check mark whether the registrant has filed a report on and attestation to its management’s assessment of the effectiveness of its internal control over financial<br>reporting under Section 404(b) of the Sarbanes-Oxley Act by the registered public accounting firm that prepared or issued its audit report.<br>If securities are registered pursuant to Section 12(b) of the Act, indicate by check mark whether the financial statements of the registrant included in the filing<br>reflect the correction of an error to previously issued financial statements.<br>Indicate by check mark whether any of those error corrections are restatements that required a recovery analysis of incentive-based compensation received by<br>any of the registrant’s executive officers during the relevant recovery period pursuant to §240.10D-1(b).<br>Indicate by check mark whether the registrant is a shell company (as defined in Rule 12b-2 of the Exchange Act). <br>Yes<br>No<br>The aggregate market value, as of June 30, 2023, of voting shares held by non-affiliates of the registrant was approximately $<br>34.1<br> billion.</code> |
* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
```json
{
"scale": 20.0,
"similarity_fct": "cos_sim"
}
```
### Training Hyperparameters
#### Non-Default Hyperparameters
- `eval_strategy`: steps
- `per_device_train_batch_size`: 10
- `per_device_eval_batch_size`: 10
- `num_train_epochs`: 5
- `multi_dataset_batch_sampler`: round_robin
#### All Hyperparameters
<details><summary>Click to expand</summary>
- `overwrite_output_dir`: False
- `do_predict`: False
- `eval_strategy`: steps
- `prediction_loss_only`: True
- `per_device_train_batch_size`: 10
- `per_device_eval_batch_size`: 10
- `per_gpu_train_batch_size`: None
- `per_gpu_eval_batch_size`: None
- `gradient_accumulation_steps`: 1
- `eval_accumulation_steps`: None
- `torch_empty_cache_steps`: None
- `learning_rate`: 5e-05
- `weight_decay`: 0.0
- `adam_beta1`: 0.9
- `adam_beta2`: 0.999
- `adam_epsilon`: 1e-08
- `max_grad_norm`: 1
- `num_train_epochs`: 5
- `max_steps`: -1
- `lr_scheduler_type`: linear
- `lr_scheduler_kwargs`: {}
- `warmup_ratio`: 0.0
- `warmup_steps`: 0
- `log_level`: passive
- `log_level_replica`: warning
- `log_on_each_node`: True
- `logging_nan_inf_filter`: True
- `save_safetensors`: True
- `save_on_each_node`: False
- `save_only_model`: False
- `restore_callback_states_from_checkpoint`: False
- `no_cuda`: False
- `use_cpu`: False
- `use_mps_device`: False
- `seed`: 42
- `data_seed`: None
- `jit_mode_eval`: False
- `use_ipex`: False
- `bf16`: False
- `fp16`: False
- `fp16_opt_level`: O1
- `half_precision_backend`: auto
- `bf16_full_eval`: False
- `fp16_full_eval`: False
- `tf32`: None
- `local_rank`: 0
- `ddp_backend`: None
- `tpu_num_cores`: None
- `tpu_metrics_debug`: False
- `debug`: []
- `dataloader_drop_last`: False
- `dataloader_num_workers`: 0
- `dataloader_prefetch_factor`: None
- `past_index`: -1
- `disable_tqdm`: False
- `remove_unused_columns`: True
- `label_names`: None
- `load_best_model_at_end`: False
- `ignore_data_skip`: False
- `fsdp`: []
- `fsdp_min_num_params`: 0
- `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
- `fsdp_transformer_layer_cls_to_wrap`: None
- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
- `deepspeed`: None
- `label_smoothing_factor`: 0.0
- `optim`: adamw_torch
- `optim_args`: None
- `adafactor`: False
- `group_by_length`: False
- `length_column_name`: length
- `ddp_find_unused_parameters`: None
- `ddp_bucket_cap_mb`: None
- `ddp_broadcast_buffers`: False
- `dataloader_pin_memory`: True
- `dataloader_persistent_workers`: False
- `skip_memory_metrics`: True
- `use_legacy_prediction_loop`: False
- `push_to_hub`: False
- `resume_from_checkpoint`: None
- `hub_model_id`: None
- `hub_strategy`: every_save
- `hub_private_repo`: False
- `hub_always_push`: False
- `gradient_checkpointing`: False
- `gradient_checkpointing_kwargs`: None
- `include_inputs_for_metrics`: False
- `eval_do_concat_batches`: True
- `fp16_backend`: auto
- `push_to_hub_model_id`: None
- `push_to_hub_organization`: None
- `mp_parameters`:
- `auto_find_batch_size`: False
- `full_determinism`: False
- `torchdynamo`: None
- `ray_scope`: last
- `ddp_timeout`: 1800
- `torch_compile`: False
- `torch_compile_backend`: None
- `torch_compile_mode`: None
- `dispatch_batches`: None
- `split_batches`: None
- `include_tokens_per_second`: False
- `include_num_input_tokens_seen`: False
- `neftune_noise_alpha`: None
- `optim_target_modules`: None
- `batch_eval_metrics`: False
- `eval_on_start`: False
- `use_liger_kernel`: False
- `eval_use_gather_object`: False
- `batch_sampler`: batch_sampler
- `multi_dataset_batch_sampler`: round_robin
</details>
### Training Logs
| Epoch | Step | cosine_map@100 |
|:------:|:----:|:--------------:|
| 0.0532 | 50 | 0.7017 |
| 0.1064 | 100 | 0.7136 |
| 0.1596 | 150 | 0.7155 |
| 0.2128 | 200 | 0.7228 |
| 0.2660 | 250 | 0.7403 |
### Framework Versions
- Python: 3.10.12
- Sentence Transformers: 3.2.1
- Transformers: 4.45.1
- PyTorch: 2.1.0+cu118
- Accelerate: 1.0.1
- Datasets: 3.1.0
- Tokenizers: 0.20.1
## Citation
### BibTeX
#### Sentence Transformers
```bibtex
@inproceedings{reimers-2019-sentence-bert,
title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
author = "Reimers, Nils and Gurevych, Iryna",
booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
month = "11",
year = "2019",
publisher = "Association for Computational Linguistics",
url = "https://arxiv.org/abs/1908.10084",
}
```
#### MultipleNegativesRankingLoss
```bibtex
@misc{henderson2017efficient,
title={Efficient Natural Language Response Suggestion for Smart Reply},
author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
year={2017},
eprint={1705.00652},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
<!--
## Glossary
*Clearly define terms in order to be accessible across audiences.*
-->
<!--
## Model Card Authors
*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
-->
<!--
## Model Card Contact
*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
--> |
hafez97/Aug-Persian-Offensive-Language-Detection | hafez97 | 2024-11-13T13:41:28Z | 107 | 0 | transformers | [
"transformers",
"safetensors",
"bert",
"text-classification",
"arxiv:1910.09700",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | text-classification | 2024-11-13T13:40:59Z | ---
library_name: transformers
tags: []
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
RichardErkhov/ArmurAI_-_Pentest_AI-gguf | RichardErkhov | 2024-11-13T13:40:52Z | 163 | 2 | null | [
"gguf",
"endpoints_compatible",
"region:us",
"conversational"
] | null | 2024-11-13T09:43:04Z | Quantization made by Richard Erkhov.
[Github](https://github.com/RichardErkhov)
[Discord](https://discord.gg/pvy7H8DZMG)
[Request more models](https://github.com/RichardErkhov/quant_request)
Pentest_AI - GGUF
- Model creator: https://huggingface.co/ArmurAI/
- Original model: https://huggingface.co/ArmurAI/Pentest_AI/
| Name | Quant method | Size |
| ---- | ---- | ---- |
| [Pentest_AI.Q2_K.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q2_K.gguf) | Q2_K | 2.53GB |
| [Pentest_AI.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q3_K_S.gguf) | Q3_K_S | 2.95GB |
| [Pentest_AI.Q3_K.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q3_K.gguf) | Q3_K | 3.28GB |
| [Pentest_AI.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q3_K_M.gguf) | Q3_K_M | 3.28GB |
| [Pentest_AI.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q3_K_L.gguf) | Q3_K_L | 3.56GB |
| [Pentest_AI.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.IQ4_XS.gguf) | IQ4_XS | 3.67GB |
| [Pentest_AI.Q4_0.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q4_0.gguf) | Q4_0 | 3.83GB |
| [Pentest_AI.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.IQ4_NL.gguf) | IQ4_NL | 3.87GB |
| [Pentest_AI.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q4_K_S.gguf) | Q4_K_S | 3.86GB |
| [Pentest_AI.Q4_K.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q4_K.gguf) | Q4_K | 4.07GB |
| [Pentest_AI.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q4_K_M.gguf) | Q4_K_M | 4.07GB |
| [Pentest_AI.Q4_1.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q4_1.gguf) | Q4_1 | 4.24GB |
| [Pentest_AI.Q5_0.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q5_0.gguf) | Q5_0 | 4.65GB |
| [Pentest_AI.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q5_K_S.gguf) | Q5_K_S | 4.65GB |
| [Pentest_AI.Q5_K.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q5_K.gguf) | Q5_K | 4.78GB |
| [Pentest_AI.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q5_K_M.gguf) | Q5_K_M | 4.78GB |
| [Pentest_AI.Q5_1.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q5_1.gguf) | Q5_1 | 5.07GB |
| [Pentest_AI.Q6_K.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q6_K.gguf) | Q6_K | 5.53GB |
| [Pentest_AI.Q8_0.gguf](https://huggingface.co/RichardErkhov/ArmurAI_-_Pentest_AI-gguf/blob/main/Pentest_AI.Q8_0.gguf) | Q8_0 | 7.17GB |
Original model description:
---
base_model: mistralai/Mistral-7B-v0.1
tags:
- mistral
- instruct
- finetune
- chatml
- gpt4
- synthetic data
- distillation
model-index:
- name: pentest ai
results: []
language:
- en
---
# Pentest AI
PentestAI is an innovative assistant for penetration testing, we used the `OpenHermes-2.5-Mistral-7B` model, we jailbroke it, finetuned it with commands for popular Kali Linux tools and it's now able to provide guided, actionable steps and command automation for performing deep pen tests.
The innovative PentestAI offers a cutting-edge solution for penetration testing by leveraging the modified OpenHermes-2.5-Mistral-7B model. This model has been uniquely jailbroken and finetuned with commands tailored for the most commonly used tools in Kali Linux, enabling it to provide guided, actionable steps and automate command execution for comprehensive penetration testing.
<img src="https://github.com/Armur-Ai/Auto-Pentest-GPT-AI/blob/main/images/Auto_Pentest.png" alt="Pentest AI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
### Key Features of PentestAI:
- **Guided Penetration Testing**: PentestAI simplifies the complexity of penetration testing by guiding you through each step of the process. Starting with the acquisition of the target IP, it offers customized advice tailored to the specific phase of the penetration test you are in.
- **Command Automation**: Streamline your penetration testing with automated command execution. PentestAI incorporates extensive knowledge of Kali Linux tools, providing you with command examples that you can execute directly or modify as needed.
- **Adaptive Learning**: As you progress through your penetration testing tasks and share results, PentestAI dynamically adapts its suggestions to enhance your efficiency and effectiveness.
- **Ethical Framework**: Throughout the testing process, PentestAI emphasizes adherence to ethical standards, ensuring that your penetration testing practices are responsible and legally compliant.
- **User-Friendly Interaction**: You can interact seamlessly with PentestAI and conclude your session anytime by typing 'exit' or 'hacked' upon successfully compromising the target machine.
By integrating these features, PentestAI not only facilitates a smoother penetration testing experience but also ensures that it is performed efficiently and ethically. Whether you are a novice eager to learn the ropes or an experienced professional seeking to streamline your workflow, PentestAI provides valuable support tailored to your needs.
For more details and to access the tool, visit the [GitHub repository](https://github.com/Armur-Ai/Auto-Pentest-GPT-AI/).
|
scilons/roberta-base-512-110k-steps-texts_pq_3-deduped-Eng_Latn | scilons | 2024-11-13T13:31:04Z | 125 | 0 | transformers | [
"transformers",
"pytorch",
"roberta",
"fill-mask",
"arxiv:1910.09700",
"autotrain_compatible",
"endpoints_compatible",
"region:us"
] | fill-mask | 2024-11-13T07:47:16Z | ---
library_name: transformers
tags: []
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
manifesto-project/manifestoberta-xlm-roberta-56policy-topics-context-2024-1-1 | manifesto-project | 2024-11-13T13:28:36Z | 184 | 3 | null | [
"pytorch",
"safetensors",
"xlm-roberta",
"custom_code",
"license:bigscience-openrail-m",
"region:us"
] | null | 2024-09-04T12:04:24Z | ---
license: bigscience-openrail-m
---
## Model description
An xlm-roberta-large model fine-tuned on ~1,7 million annotated statements contained in the [Manifesto Corpus](https://manifesto-project.wzb.eu/information/documents/corpus) (version 2024a).
The model can be used to categorize any type of text into 56 different political topics according to the Manifesto Project's coding scheme ([Handbook 4](https://manifesto-project.wzb.eu/coding_schemes/mp_v4)).
It works for all languages the xlm-roberta model is pretrained on ([overview](https://github.com/facebookresearch/fairseq/tree/main/examples/xlmr#introduction)), just note that it will perform best for the 38 languages contained in the Manifesto Corpus:
||||||
|------|------|------|------|------|
|armenian|bosnian|bulgarian|catalan|croatian|
|czech|danish|dutch|english|estonian|
|finnish|french|galician|georgian|german|
|greek|hebrew|hungarian|icelandic|italian|
|japanese|korean|latvian|lithuanian|macedonian|
|montenegrin|norwegian|polish|portuguese|romanian|
|russian|serbian|slovak|slovenian|spanish|
|swedish|turkish|ukrainian| | |
The context model variant additionally incorporates the surrounding sentences of a statement to improve the classification results for ambiguous sentences. (See Training Procedure for details)
**Important**
We slightly modified the Classification Head of the `XLMRobertaModelForSequenceClassification` model (removed the tanh activation and the intermediate linear layer) as that improved the model performance for this task considerably.
To correctly load the full model, include the `trust_remote_code=True` argument when using the `from_pretrained method`.
## How to use
```python
import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer
model = AutoModelForSequenceClassification.from_pretrained("manifesto-project/manifestoberta-xlm-roberta-56policy-topics-context-2024-1-1", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large")
sentence = "These principles are under threat."
context = "Human rights and international humanitarian law are fundamental pillars of a secure global system. These principles are under threat. Some of the world's most powerful states choose to sell arms to human-rights abusing states."
# For sentences without additional context, just use the sentence itself as the context.
# Example: context = "These principles are under threat."
inputs = tokenizer(sentence,
context,
return_tensors="pt",
max_length=300, #we limited the input to 300 tokens during finetuning
padding="max_length",
truncation=True
)
logits = model(**inputs).logits
probabilities = torch.softmax(logits, dim=1).tolist()[0]
probabilities = {model.config.id2label[index]: round(probability * 100, 2) for index, probability in enumerate(probabilities)}
probabilities = dict(sorted(probabilities.items(), key=lambda item: item[1], reverse=True))
print(probabilities)
# {'201 - Freedom and Human Rights': 86.58, '107 - Internationalism: Positive': 8.12, '106 - Peace': 1.67...
predicted_class = model.config.id2label[logits.argmax().item()]
print(predicted_class)
# 201 - Freedom and Human Rights
```
## Training Procedure
Training of the model took place on all quasi-sentences of the Manifesto Corpus (version 2024a), minus 10% that were kept out of training for the final test and evaluation results.
This results in a training dataset of 1,691,707 quasi-sentences.
As our context-including model input poses the threat of data-leakage problems between train and test data, we refrained from randomly splitting quasi-sentences into train and test data.
Instead, we randomly split the dataset on the manifesto level, so that 1828 manifestos and all their quasi-sentences were assigned to the train set and 203 to the test set.
As training parameters, we used the following settings: learning rate: 1e-5, weight decay: 0.01, epochs: 1, batch size: 4, gradient accumulation steps: 4 (effective batch size: 16).
### Context
To adapt the model to the task of classifying statements in manifestos we made some modifications to the traditional training setup.
Given that human annotators in the Manifesto Project are encouraged to use surrounding sentences to interpret ambiguous statements , we combined statements with their context for our model's input.
Specifically, we used a sentence-pair input, where the single to-be-classified statement gets followed by the separator token followed by the whole bigger context of length 200 tokens, in which the statement to-be-classified is embedded.
Here is an example:
*"`<s>` We must right the wrongs in our democracy, `</s>` `</s>` To turn this crisis into a crucible, from which we will forge a stronger, brighter, and more equitable future. We must right the wrongs in our democracy, redress the systemic injustices that have long plagued our society,throw open the doors of opportunity for all Americans and reinvent our institutions at home and our leadership abroad. `</s>`".*
The second part, which contains the context, is greedily filled until it contains 200 tokens.
Our tests showed that including the context helped to improve the performance of the classification model considerably (~7% accuracy).
We tried other approaches like using two XLM-RoBERTa models as a duo, where one receives the sentence and one the context, and a shared-layer model, where both inputs are fed separately trough the same model.
Both variants performed similarly to our sentence pair approach, but lead to higher complexity and computing costs, which is why we ultimately opted for the sentence pair way to include the surrounding context.
## Model Performance
The model was evaluated on a test set of 200,920 annotated manifesto statements.
### Overall
| | Accuracy | Top2_Acc | Top3_Acc | Precision| Recall | F1_Macro | MCC | Cross-Entropy |
|-------------------------------------------------------------------------------------------------------|:--------:|:--------:|:--------:|:--------:|:------:|:--------:|:---:|:-------------:|
[Sentence Model](https://huggingface.co/manifesto-project/manifestoberta-xlm-roberta-56policy-topics-sentence-2024-1-1)| 0.57 | 0.73 | 0.81 | 0.48 | 0.43 | 0.45 | 0.55| 1.47 |
[Context Model](https://huggingface.co/manifesto-project/manifestoberta-xlm-roberta-56policy-topics-context-2024-1-1) | 0.64 | 0.81 | 0.88 | 0.55 | 0.52 | 0.53 | 0.63| 1.15 |
## Citation
Please cite the model as follows:
Burst, Tobias / Lehmann, Pola / Franzmann, Simon / Al-Gaddooa, Denise / Ivanusch, Christoph / Regel, Sven / Riethmüller, Felicia / Weßels, Bernhard / Zehnter, Lisa (2024): manifestoberta. Version 56topics.context.2024.1.1. Berlin: Wissenschaftszentrum Berlin für Sozialforschung (WZB) / Göttingen: Institut für Demokratieforschung (IfDem). https://doi.org/10.25522/manifesto.manifestoberta.56topics.context.2024.1.1
```bib
@misc{Burst:2024,
Address = {Berlin / Göttingen},
Author = {Burst, Tobias AND Lehmann, Pola AND Franzmann, Simon AND Al-Gaddooa, Denise AND Ivanusch, Christoph AND Regel, Sven AND Riethmüller, Felicia AND Weßels, Bernhard AND Zehnter, Lisa},
Publisher = {Wissenschaftszentrum Berlin für Sozialforschung / Göttinger Institut für Demokratieforschung},
Title = {manifestoberta. Version 56topics.context.2024.1.1},
doi = {10.25522/manifesto.manifestoberta.56topics.context.2024.1.1},
url = {https://doi.org/10.25522/manifesto.manifestoberta.56topics.context.2024.1.1},
Year = {2024},
``` |
PrunaAI/resnet152.tv2_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:41Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:25:33Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet152.tv2_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet152.tv2_in1k-turbo-green-smashed --local-dir resnet152.tv2_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet152.tv2_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet152.tv2_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet152.tv2_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:40Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T16:36:18Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed --local-dir convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_base.clip_laiona_augreg_320-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_base.clip_laiona_augreg_320 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:36Z | 4 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:44:02Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed --local-dir convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_base.fb_in22k_ft_in1k_384-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_base.fb_in22k_ft_in1k_384 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet34.bt_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:35Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:54:15Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet34.bt_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet34.bt_in1k-turbo-green-smashed --local-dir resnet34.bt_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet34.bt_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet34.bt_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet34.bt_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50.b1k_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:34Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:59:18Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50.b1k_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50.b1k_in1k-turbo-green-smashed --local-dir resnet50.b1k_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50.b1k_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50.b1k_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50.b1k_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:33Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:11:08Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed --local-dir maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_rmlp_base_rw_224.sw_in12k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:32Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:14:40Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed --local-dir vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_huge_patch14_clip_224.laion2b_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_huge_patch14_clip_224.laion2b_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:30Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:30:18Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed --local-dir resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnext50_32x4d.fb_swsl_ig1b_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnext50_32x4d.fb_swsl_ig1b_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet101c.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:30Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:17:13Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet101c.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet101c.gluon_in1k-turbo-green-smashed --local-dir resnet101c.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet101c.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet101c.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet101c.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
manifesto-project/manifestoberta-xlm-roberta-56policy-topics-sentence-2024-1-1 | manifesto-project | 2024-11-13T13:25:26Z | 20 | 1 | null | [
"pytorch",
"safetensors",
"xlm-roberta",
"custom_code",
"license:bigscience-openrail-m",
"region:us"
] | null | 2024-09-04T11:46:07Z | ---
license: bigscience-openrail-m
widget:
- text: >-
We will restore funding to the Global Environment Facility and the
Intergovernmental Panel on Climate Change.
---
## Model description
An xlm-roberta-large model fine-tuned on ~1,7 million annotated statements contained in the [Manifesto Corpus](https://manifesto-project.wzb.eu/information/documents/corpus) (version 2024a).
The model can be used to categorize any type of text into 56 different political topics according to the Manifesto Project's coding scheme ([Handbook 4](https://manifesto-project.wzb.eu/coding_schemes/mp_v4)).
It works for all languages the xlm-roberta model is pretrained on ([overview](https://github.com/facebookresearch/fairseq/tree/main/examples/xlmr#introduction)), just note that it will perform best for the 38 languages contained in the Manifesto Corpus:
||||||
|------|------|------|------|------|
|armenian|bosnian|bulgarian|catalan|croatian|
|czech|danish|dutch|english|estonian|
|finnish|french|galician|georgian|german|
|greek|hebrew|hungarian|icelandic|italian|
|japanese|korean|latvian|lithuanian|macedonian|
|montenegrin|norwegian|polish|portuguese|romanian|
|russian|serbian|slovak|slovenian|spanish|
|swedish|turkish|ukrainian| | |
## How to use
```python
import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer
model = AutoModelForSequenceClassification.from_pretrained("manifesto-project/manifestoberta-xlm-roberta-56policy-topics-sentence-2024-1-1")
tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large")
sentence = "We will restore funding to the Global Environment Facility and the Intergovernmental Panel on Climate Change, to support critical climate science research around the world"
inputs = tokenizer(sentence,
return_tensors="pt",
max_length=200, #we limited the input to 200 tokens during finetuning
padding="max_length",
truncation=True
)
logits = model(**inputs).logits
probabilities = torch.softmax(logits, dim=1).tolist()[0]
probabilities = {model.config.id2label[index]: round(probability * 100, 2) for index, probability in enumerate(probabilities)}
probabilities = dict(sorted(probabilities.items(), key=lambda item: item[1], reverse=True))
print(probabilities)
# {'501 - Environmental Protection: Positive': 67.56, '411 - Technology and Infrastructure': 14.03, '107 - Internationalism: Positive': 13.58, '416 - Anti-Growth Economy: Positive': 2.24...
predicted_class = model.config.id2label[logits.argmax().item()]
print(predicted_class)
# 501 - Environmental Protection: Positive
```
## Model Performance
The model was evaluated on a test set of 200,920 annotated manifesto statements.
### Overall
| | Accuracy | Top2_Acc | Top3_Acc | Precision| Recall | F1_Macro | MCC | Cross-Entropy |
|-------------------------------------------------------------------------------------------------------|:--------:|:--------:|:--------:|:--------:|:------:|:--------:|:---:|:-------------:|
[Sentence Model](https://huggingface.co/manifesto-project/manifestoberta-xlm-roberta-56policy-topics-sentence-2024-1-1)| 0.57 | 0.73 | 0.81 | 0.48 | 0.43 | 0.45 | 0.55| 1.47 |
[Context Model](https://huggingface.co/manifesto-project/manifestoberta-xlm-roberta-56policy-topics-context-2024-1-1) | 0.64 | 0.81 | 0.88 | 0.55 | 0.52 | 0.53 | 0.63| 1.15 |
### Citation
Please cite the model as follows:
Burst, Tobias / Lehmann, Pola / Franzmann, Simon / Al-Gaddooa, Denise / Ivanusch, Christoph / Regel, Sven / Riethmüller, Felicia / Weßels, Bernhard / Zehnter, Lisa (2024): manifestoberta. Version 56topics.sentence.2024.1.1. Berlin: Wissenschaftszentrum Berlin für Sozialforschung (WZB) / Göttingen: Institut für Demokratieforschung (IfDem). https://doi.org/10.25522/manifesto.manifestoberta.56topics.sentence.2024.1.1
```bib
@misc{Burst:2024,
Address = {Berlin / Göttingen},
Author = {Burst, Tobias AND Lehmann, Pola AND Franzmann, Simon AND Al-Gaddooa, Denise AND Ivanusch, Christoph AND Regel, Sven AND Riethmüller, Felicia AND Weßels, Bernhard AND Zehnter, Lisa},
Publisher = {Wissenschaftszentrum Berlin für Sozialforschung / Göttinger Institut für Demokratieforschung},
Title = {manifestoberta. Version 56topics.sentence.2024.1.1},
doi = {10.25522/manifesto.manifestoberta.56topics.sentence.2024.1.1},
url = {https://doi.org/10.25522/manifesto.manifestoberta.56topics.sentence.2024.1.1},
Year = {2024},
``` |
PrunaAI/wide_resnet50_2.racm_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:24Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:38:41Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir wide_resnet50_2.racm_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/wide_resnet50_2.racm_in1k-turbo-green-smashed --local-dir wide_resnet50_2.racm_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "wide_resnet50_2.racm_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "wide_resnet50_2.racm_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model wide_resnet50_2.racm_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:22Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:38:46Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed --local-dir convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_xxlarge.clip_laion2b_soup-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_xxlarge.clip_laion2b_soup before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet34.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:21Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:54:46Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet34.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet34.gluon_in1k-turbo-green-smashed --local-dir resnet34.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet34.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet34.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet34.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
mahbubhimel/BNGEMMA | mahbubhimel | 2024-11-13T13:25:21Z | 91 | 0 | transformers | [
"transformers",
"safetensors",
"gemma2",
"text-classification",
"arxiv:1910.09700",
"autotrain_compatible",
"text-generation-inference",
"endpoints_compatible",
"region:us"
] | text-classification | 2024-11-13T13:22:11Z | ---
library_name: transformers
tags: []
---
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed] |
PrunaAI/gcvit_tiny.in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:20Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T19:18:40Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir gcvit_tiny.in1k-turbo-green-smashed
huggingface-cli download PrunaAI/gcvit_tiny.in1k-turbo-green-smashed --local-dir gcvit_tiny.in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "gcvit_tiny.in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "gcvit_tiny.in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model gcvit_tiny.in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:19Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T10:00:14Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed --local-dir seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "seresnextaa101d_32x8d.ah_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model seresnextaa101d_32x8d.ah_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:18Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T10:48:33Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed --local-dir tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "tresnet_v2_l.miil_in21k_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model tresnet_v2_l.miil_in21k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:17Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:06:45Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed --local-dir vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_base_patch32_clip_224.laion2b-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_base_patch32_clip_224.laion2b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50d.a2_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:16Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:07:29Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50d.a2_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50d.a2_in1k-turbo-green-smashed --local-dir resnet50d.a2_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50d.a2_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50d.a2_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50d.a2_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:16Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T10:06:09Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed --local-dir convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convformer_s18.sail_in22k_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convformer_s18.sail_in22k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:15Z | 3 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:29:15Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed --local-dir vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_large_patch14_clip_336.openai-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_large_patch14_clip_336.openai before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:14Z | 3 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-09T04:48:17Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed --local-dir convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_large_mlp.clip_laion2b_soup_ft_in12k_320-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_large_mlp.clip_laion2b_soup_ft_in12k_320 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50d.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:13Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:07:54Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50d.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50d.gluon_in1k-turbo-green-smashed --local-dir resnet50d.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50d.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50d.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50d.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilenetv2_100.ra_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:12Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:40:45Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilenetv2_100.ra_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/mobilenetv2_100.ra_in1k-turbo-green-smashed --local-dir mobilenetv2_100.ra_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilenetv2_100.ra_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilenetv2_100.ra_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilenetv2_100.ra_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:11Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:50:59Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed --local-dir resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetv2_50x1_bit.goog_in21k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetv2_50x1_bit.goog_in21k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet26.bt_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:07Z | 5 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:50:25Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet26.bt_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet26.bt_in1k-turbo-green-smashed --local-dir resnet26.bt_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet26.bt_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet26.bt_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet26.bt_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_nano.fcmae-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:06Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:49:11Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_nano.fcmae-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_nano.fcmae-turbo-tiny-green-smashed --local-dir convnextv2_nano.fcmae-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_nano.fcmae-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_nano.fcmae-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_nano.fcmae before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:04Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T11:46:05Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed --local-dir maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_rmlp_pico_rw_256.sw_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_rmlp_pico_rw_256.sw_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/seresnext101d_32x8d.ah_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:03Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:59:37Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir seresnext101d_32x8d.ah_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/seresnext101d_32x8d.ah_in1k-turbo-green-smashed --local-dir seresnext101d_32x8d.ah_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "seresnext101d_32x8d.ah_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "seresnext101d_32x8d.ah_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model seresnext101d_32x8d.ah_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevitv2_175.cvnets_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:02Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:59:37Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevitv2_175.cvnets_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevitv2_175.cvnets_in1k-turbo-green-smashed --local-dir mobilevitv2_175.cvnets_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevitv2_175.cvnets_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevitv2_175.cvnets_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevitv2_175.cvnets_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetrs270.tf_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:25:01Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:44:04Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetrs270.tf_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetrs270.tf_in1k-turbo-green-smashed --local-dir resnetrs270.tf_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetrs270.tf_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetrs270.tf_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetrs270.tf_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_huge.fcmae-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:25:00Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:48:46Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_huge.fcmae-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_huge.fcmae-turbo-tiny-green-smashed --local-dir convnextv2_huge.fcmae-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_huge.fcmae-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_huge.fcmae-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_huge.fcmae before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:59Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:11:57Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed --local-dir vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_large_r50_s32_384.augreg_in21k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_large_r50_s32_384.augreg_in21k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:58Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:00:23Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed --local-dir maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_base_tf_384.in21k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_base_tf_384.in21k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convformer_b36.sail_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:57Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T09:52:41Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convformer_b36.sail_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/convformer_b36.sail_in1k-turbo-green-smashed --local-dir convformer_b36.sail_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convformer_b36.sail_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convformer_b36.sail_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convformer_b36.sail_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet18-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:55Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T13:38:04Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet18-turbo-green-smashed
huggingface-cli download PrunaAI/resnet18-turbo-green-smashed --local-dir resnet18-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet18-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet18-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet18 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:54Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-09T05:20:54Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed --local-dir convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_xlarge.fb_in22k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_xlarge.fb_in22k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:53Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:59:05Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed --local-dir mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevitv2_150.cvnets_in22k_ft_in1k_384-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevitv2_150.cvnets_in22k_ft_in1k_384 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:51Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:45:57Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed --local-dir resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnext101_32x32d.fb_wsl_ig1b_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnext101_32x32d.fb_wsl_ig1b_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/flexivit_small.300ep_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:48Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T11:35:08Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir flexivit_small.300ep_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/flexivit_small.300ep_in1k-turbo-tiny-green-smashed --local-dir flexivit_small.300ep_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "flexivit_small.300ep_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "flexivit_small.300ep_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model flexivit_small.300ep_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:48Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T16:31:41Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_base.clip_laion2b_augreg_ft_in12k_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_base.clip_laion2b_augreg_ft_in12k_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetrs200.tf_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:43Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:41:30Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetrs200.tf_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetrs200.tf_in1k-turbo-green-smashed --local-dir resnetrs200.tf_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetrs200.tf_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetrs200.tf_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetrs200.tf_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_base-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:42Z | 3 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T13:08:14Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_base-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_base-turbo-tiny-green-smashed --local-dir convnext_base-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_base-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_base-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_base before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:40Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:06:47Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed --local-dir convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_large_mlp.clip_laion2b_ft_soup_320-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_large_mlp.clip_laion2b_ft_soup_320 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetrs101.tf_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:40Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:36:54Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetrs101.tf_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetrs101.tf_in1k-turbo-green-smashed --local-dir resnetrs101.tf_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetrs101.tf_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetrs101.tf_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetrs101.tf_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_tiny.fb_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:39Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:26:21Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_tiny.fb_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_tiny.fb_in1k-turbo-tiny-green-smashed --local-dir convnext_tiny.fb_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_tiny.fb_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_tiny.fb_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_tiny.fb_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:38Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:36:17Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed --local-dir resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnext101_32x4d.fb_ssl_yfcc100m_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50s.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:37Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:10:02Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50s.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50s.gluon_in1k-turbo-green-smashed --local-dir resnet50s.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50s.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50s.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50s.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevitv2_125.cvnets_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:35Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:56:02Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevitv2_125.cvnets_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevitv2_125.cvnets_in1k-turbo-green-smashed --local-dir mobilevitv2_125.cvnets_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevitv2_125.cvnets_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevitv2_125.cvnets_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevitv2_125.cvnets_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:34Z | 4 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:00:20Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed --local-dir vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_base_patch16_224.sam_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_base_patch16_224.sam_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:28Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:08:26Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed --local-dir vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_large_patch32_384.orig_in21k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_large_patch32_384.orig_in21k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/flexivit_base.300ep_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:28Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T11:28:45Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir flexivit_base.300ep_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/flexivit_base.300ep_in1k-turbo-tiny-green-smashed --local-dir flexivit_base.300ep_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "flexivit_base.300ep_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "flexivit_base.300ep_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model flexivit_base.300ep_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:24Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:41:39Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed
huggingface-cli download PrunaAI/convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed --local-dir convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convformer_s18.sail_in22k_ft_in1k_384-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convformer_s18.sail_in22k_ft_in1k_384 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:22Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T06:02:05Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed --local-dir mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevitv2_175.cvnets_in22k_ft_in1k_384-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevitv2_175.cvnets_in22k_ft_in1k_384 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:20Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:34:53Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed --local-dir convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_xxlarge.clip_laion2b_rewind-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_xxlarge.clip_laion2b_rewind before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevitv2_050.cvnets_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:19Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:53:26Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevitv2_050.cvnets_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevitv2_050.cvnets_in1k-turbo-green-smashed --local-dir mobilevitv2_050.cvnets_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevitv2_050.cvnets_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevitv2_050.cvnets_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevitv2_050.cvnets_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:16Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:04:08Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed --local-dir vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_base_patch16_384.augreg_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_base_patch16_384.augreg_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:15Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:24:10Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed --local-dir convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_small.fb_in22k_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_small.fb_in22k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_base.fcmae-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:13Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:39:56Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_base.fcmae-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_base.fcmae-turbo-tiny-green-smashed --local-dir convnextv2_base.fcmae-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_base.fcmae-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_base.fcmae-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_base.fcmae before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:13Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:09:45Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed --local-dir maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_rmlp_base_rw_224.sw_in12k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_rmlp_base_rw_224.sw_in12k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet34.a2_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:12Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:52:44Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet34.a2_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet34.a2_in1k-turbo-green-smashed --local-dir resnet34.a2_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet34.a2_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet34.a2_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet34.a2_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:11Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:20:38Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_large_patch14_clip_224.laion2b_ft_in12k_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_large_patch14_clip_224.laion2b_ft_in12k_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_large-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:10Z | 3 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T13:12:20Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_large-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnext_large-turbo-tiny-green-smashed --local-dir convnext_large-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_large-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_large-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_large before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetrs152.tf_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:09Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:38:37Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetrs152.tf_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetrs152.tf_in1k-turbo-green-smashed --local-dir resnetrs152.tf_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetrs152.tf_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetrs152.tf_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetrs152.tf_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_base_tf_512.in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:24:04Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:33:57Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_base_tf_512.in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/maxvit_base_tf_512.in1k-turbo-tiny-green-smashed --local-dir maxvit_base_tf_512.in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_base_tf_512.in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_base_tf_512.in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_base_tf_512.in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/tresnet_m.miil_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:24:00Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T10:45:33Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir tresnet_m.miil_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/tresnet_m.miil_in1k-turbo-green-smashed --local-dir tresnet_m.miil_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "tresnet_m.miil_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "tresnet_m.miil_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model tresnet_m.miil_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/seresnet50.ra2_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:58Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:54:11Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir seresnet50.ra2_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/seresnet50.ra2_in1k-turbo-green-smashed --local-dir seresnet50.ra2_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "seresnet50.ra2_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "seresnet50.ra2_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model seresnet50.ra2_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:57Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T13:34:58Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed --local-dir convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_huge.fcmae_ft_in22k_in1k_512-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_huge.fcmae_ft_in22k_in1k_512 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/maxvit_t-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:56Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T13:24:31Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir maxvit_t-turbo-green-smashed
huggingface-cli download PrunaAI/maxvit_t-turbo-green-smashed --local-dir maxvit_t-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "maxvit_t-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "maxvit_t-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model maxvit_t before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:56Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:51:52Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed --local-dir convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_huge.fcmae_ft_in22k_in1k_384-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_huge.fcmae_ft_in22k_in1k_384 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:52Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:52:54Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed --local-dir convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnextv2_tiny.fcmae_ft_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnextv2_tiny.fcmae_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_femto.d1_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:50Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T16:41:37Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_femto.d1_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/convnext_femto.d1_in1k-turbo-green-smashed --local-dir convnext_femto.d1_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_femto.d1_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_femto.d1_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_femto.d1_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50c.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:49Z | 3 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:06:25Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50c.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50c.gluon_in1k-turbo-green-smashed --local-dir resnet50c.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50c.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50c.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50c.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet34-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:48Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T13:38:29Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet34-turbo-green-smashed
huggingface-cli download PrunaAI/resnet34-turbo-green-smashed --local-dir resnet34-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet34-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet34-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet34 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:47Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:35:41Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed --local-dir vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_srelpos_small_patch16_224.sw_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_srelpos_small_patch16_224.sw_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet50.a3_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:47Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:57:51Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet50.a3_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet50.a3_in1k-turbo-green-smashed --local-dir resnet50.a3_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet50.a3_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet50.a3_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet50.a3_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/convnext_tiny_hnf.a2h_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:46Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T17:29:09Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir convnext_tiny_hnf.a2h_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/convnext_tiny_hnf.a2h_in1k-turbo-green-smashed --local-dir convnext_tiny_hnf.a2h_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "convnext_tiny_hnf.a2h_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "convnext_tiny_hnf.a2h_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model convnext_tiny_hnf.a2h_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:44Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T09:33:10Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed --local-dir resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetaa101d.sw_in12k_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetaa101d.sw_in12k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:41Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T11:05:52Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed --local-dir vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_base_patch32_224.augreg_in21k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_base_patch32_224.augreg_in21k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/gcresnext50ts.ch_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:38Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-07T19:14:20Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir gcresnext50ts.ch_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/gcresnext50ts.ch_in1k-turbo-green-smashed --local-dir gcresnext50ts.ch_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "gcresnext50ts.ch_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "gcresnext50ts.ch_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model gcresnext50ts.ch_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:36Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:30:07Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_large_patch14_clip_336.openai_ft_in12k_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_large_patch14_clip_336.openai_ft_in12k_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:35Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:21:50Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed --local-dir resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnetv2_152x2_bit.goog_teacher_in21k_ft_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnext50_32x4d.gluon_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:34Z | 4 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-14T09:31:38Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnext50_32x4d.gluon_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnext50_32x4d.gluon_in1k-turbo-green-smashed --local-dir resnext50_32x4d.gluon_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnext50_32x4d.gluon_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnext50_32x4d.gluon_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnext50_32x4d.gluon_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/resnet26t.ra2_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:33Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T08:51:09Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir resnet26t.ra2_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/resnet26t.ra2_in1k-turbo-green-smashed --local-dir resnet26t.ra2_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "resnet26t.ra2_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "resnet26t.ra2_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model resnet26t.ra2_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:33Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T12:02:07Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed --local-dir vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "vit_relpos_base_patch32_plus_rpn_256.sw_in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model vit_relpos_base_patch32_plus_rpn_256.sw_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/mobilevit_xs.cvnets_in1k-turbo-green-smashed | PrunaAI | 2024-11-13T13:23:27Z | 2 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-10T05:51:26Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir mobilevit_xs.cvnets_in1k-turbo-green-smashed
huggingface-cli download PrunaAI/mobilevit_xs.cvnets_in1k-turbo-green-smashed --local-dir mobilevit_xs.cvnets_in1k-turbo-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "mobilevit_xs.cvnets_in1k-turbo-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "mobilevit_xs.cvnets_in1k-turbo-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model mobilevit_xs.cvnets_in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
PrunaAI/crossvit_15_240.in1k-turbo-tiny-green-smashed | PrunaAI | 2024-11-13T13:23:25Z | 1 | 0 | pruna-engine | [
"pruna-engine",
"region:us"
] | null | 2024-03-19T11:24:32Z | ---
library_name: pruna-engine
thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
metrics:
- memory_disk
- memory_inference
- inference_latency
- inference_throughput
- inference_CO2_emissions
- inference_energy_consumption
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<a href="https://docs.pruna.ai/en/latest/setup/pip.html" target="_blank" rel="noopener noreferrer">
<img src="https://imgur.com/rVAgqMY.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</a>
</div>
<!-- header end -->
[](https://twitter.com/PrunaAI)
[](https://github.com/PrunaAI)
[](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
[](https://discord.gg/rskEr4BZJx)
<div style="color: #9B1DBE; font-size: 2em; font-weight: bold;">
Deprecation Notice: This model is deprecated and will no longer receive updates.
</div>
<br><br>
# Simply make AI models cheaper, smaller, faster, and greener!
- Give a thumbs up if you like this model!
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
- Join Pruna AI community on Discord [here](https://discord.gg/rskEr4BZJx) to share feedback/suggestions or get help.
## Results
**Frequently Asked Questions**
- ***How does the compression work?*** The model is compressed by combining quantization, xformers, jit, cuda graphs, triton.
- ***How does the model quality change?*** The quality of the model output might slightly vary compared to the base model.
- ***How is the model efficiency evaluated?*** These results were obtained on NVIDIA A100-PCIE-40GB with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
- ***What is the model format?*** We used a custom Pruna model format based on pickle to make models compatible with the compression methods. We provide a tutorial to run models in dockers in the documentation [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) if needed.
- ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
- ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
- ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
- ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
## Setup
You can run the smashed model with these steps:
0. Check that you have linux, python 3.10, and cuda 12.1.0 requirements installed. For cuda, check with `nvcc --version` and install with `conda install nvidia/label/cuda-12.1.0::cuda`.
1. Install the `pruna-engine` available [here](https://pypi.org/project/pruna-engine/) on Pypi. It might take up to 15 minutes to install.
```bash
pip install pruna-engine[gpu]==0.7.1 --extra-index-url https://pypi.nvidia.com --extra-index-url https://pypi.ngc.nvidia.com --extra-index-url https://prunaai.pythonanywhere.com/
```
2. Download the model files using one of these three options.
- Option 1 - Use command line interface (CLI):
```bash
mkdir crossvit_15_240.in1k-turbo-tiny-green-smashed
huggingface-cli download PrunaAI/crossvit_15_240.in1k-turbo-tiny-green-smashed --local-dir crossvit_15_240.in1k-turbo-tiny-green-smashed --local-dir-use-symlinks False
```
- Option 2 - Use Python:
```python
import subprocess
repo_name = "crossvit_15_240.in1k-turbo-tiny-green-smashed"
subprocess.run(["mkdir", repo_name])
subprocess.run(["huggingface-cli", "download", 'PrunaAI/'+ repo_name, "--local-dir", repo_name, "--local-dir-use-symlinks", "False"])
```
- Option 3 - Download them manually on the HuggingFace model page.
3. Load & run the model.
```python
from pruna_engine.PrunaModel import PrunaModel
model_path = "crossvit_15_240.in1k-turbo-tiny-green-smashed/model" # Specify the downloaded model path.
smashed_model = PrunaModel.load_model(model_path) # Load the model.
import torch; image = torch.rand(1, 3, 224, 224).to('cuda')
smashed_model(image)
```
## Configurations
The configuration info are in `model/smash_config.json`.
## Credits & License
The license of the smashed model follows the license of the original model. Please check the license of the original model crossvit_15_240.in1k before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
## Want to compress other models?
- Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
- Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). |
Subsets and Splits