modelId
stringlengths 4
112
| sha
stringlengths 40
40
| lastModified
stringlengths 24
24
| tags
sequence | pipeline_tag
stringclasses 29
values | private
bool 1
class | author
stringlengths 2
38
⌀ | config
null | id
stringlengths 4
112
| downloads
float64 0
36.8M
⌀ | likes
float64 0
712
⌀ | library_name
stringclasses 17
values | __index_level_0__
int64 0
38.5k
| readme
stringlengths 0
186k
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
TencentGameMate/chinese-hubert-base | fce0375452b1dd6c080ac3248d423d4d037bc831 | 2022-06-24T01:52:57.000Z | [
"pytorch",
"hubert",
"feature-extraction",
"transformers",
"license:mit"
] | feature-extraction | false | TencentGameMate | null | TencentGameMate/chinese-hubert-base | 953 | 3 | transformers | 1,800 | ---
license: mit
---
Pretrained on 10k hours WenetSpeech L subset. More details in [TencentGameMate/chinese_speech_pretrain](https://github.com/TencentGameMate/chinese_speech_pretrain)
This model does not have a tokenizer as it was pretrained on audio alone.
In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data.
python package:
transformers==4.16.2
```python
import torch
import torch.nn.functional as F
import soundfile as sf
from transformers import (
Wav2Vec2FeatureExtractor,
HubertModel,
)
model_path=""
wav_path=""
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(model_path)
model = HubertModel.from_pretrained(model_path)
# for pretrain: Wav2Vec2ForPreTraining
# model = Wav2Vec2ForPreTraining.from_pretrained(model_path)
model = model.to(device)
model = model.half()
model.eval()
wav, sr = sf.read(wav_path)
input_values = feature_extractor(wav, return_tensors="pt").input_values
input_values = input_values.half()
input_values = input_values.to(device)
with torch.no_grad():
outputs = model(input_values)
last_hidden_state = outputs.last_hidden_state
``` |
Norod78/hebrew-gpt_neo-small | 12ac2ca1aac05eeaab5e3bb278fd4e31180b7545 | 2022-07-04T12:43:15.000Z | [
"pytorch",
"jax",
"onnx",
"gpt_neo",
"text-generation",
"he",
"transformers",
"license:mit"
] | text-generation | false | Norod78 | null | Norod78/hebrew-gpt_neo-small | 946 | null | transformers | 1,801 | ---
language: he
thumbnail: https://avatars1.githubusercontent.com/u/3617152?norod.jpg
widget:
- text: "עוד בימי קדם"
- text: "קוראים לי דורון ואני מעוניין ל"
- text: "קוראים לי איציק ואני חושב ש"
- text: "החתול שלך מאוד חמוד ו"
license: mit
---
# hebrew-gpt_neo-small
Hebrew text generation model based on [EleutherAI's gpt-neo](https://github.com/EleutherAI/gpt-neo). Each was trained on a TPUv3-8 which was made avilable to me via the [TPU Research Cloud](https://sites.research.google/trc/) Program.
## Datasets
1. An assortment of various Hebrew corpuses - I have made it available [here](https://mega.nz/folder/CodSSA4R#4INvMes-56m_WUi7jQMbJQ)
2. oscar / unshuffled_deduplicated_he - [Homepage](https://oscar-corpus.com) | [Dataset Permalink](https://huggingface.co/datasets/viewer/?dataset=oscar&config=unshuffled_deduplicated_he)
The Open Super-large Crawled ALMAnaCH coRpus is a huge multilingual corpus obtained by language classification and filtering of the Common Crawl corpus using the goclassy architecture.
3. CC100-Hebrew Dataset [Homepage](https://metatext.io/datasets/cc100-hebrew)
Created by Conneau & Wenzek et al. at 2020, the CC100-Hebrew This dataset is one of the 100 corpora of monolingual data that was processed from the January-December 2018 Commoncrawl snapshots from the CC-Net repository. The size of this corpus is 6.1G., in Hebrew language.
## Training Config
Available [here](https://github.com/Norod/hebrew-gpt_neo/tree/main/hebrew-gpt_neo-small/configs) <BR>
## Usage
### Google Colab Notebook
Available [here ](https://colab.research.google.com/github/Norod/hebrew-gpt_neo/blob/main/hebrew-gpt_neo-small/Norod78_hebrew_gpt_neo_small_Colab.ipynb) <BR>
#### Simple usage sample code
```python
!pip install tokenizers==0.10.2 transformers==4.6.0
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("Norod78/hebrew-gpt_neo-small")
model = AutoModelForCausalLM.from_pretrained("Norod78/hebrew-gpt_neo-small", pad_token_id=tokenizer.eos_token_id)
prompt_text = "אני אוהב שוקולד ועוגות"
max_len = 512
sample_output_num = 3
seed = 1000
import numpy as np
import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
n_gpu = 0 if torch.cuda.is_available()==False else torch.cuda.device_count()
print(f"device: {device}, n_gpu: {n_gpu}")
np.random.seed(seed)
torch.manual_seed(seed)
if n_gpu > 0:
torch.cuda.manual_seed_all(seed)
model.to(device)
encoded_prompt = tokenizer.encode(
prompt_text, add_special_tokens=False, return_tensors="pt")
encoded_prompt = encoded_prompt.to(device)
if encoded_prompt.size()[-1] == 0:
input_ids = None
else:
input_ids = encoded_prompt
print("input_ids = " + str(input_ids))
if input_ids != None:
max_len += len(encoded_prompt[0])
if max_len > 2048:
max_len = 2048
print("Updated max_len = " + str(max_len))
stop_token = "<|endoftext|>"
new_lines = "\n\n\n"
sample_outputs = model.generate(
input_ids,
do_sample=True,
max_length=max_len,
top_k=50,
top_p=0.95,
num_return_sequences=sample_output_num
)
print(100 * '-' + "\n\t\tOutput\n" + 100 * '-')
for i, sample_output in enumerate(sample_outputs):
text = tokenizer.decode(sample_output, skip_special_tokens=True)
# Remove all text after the stop token
text = text[: text.find(stop_token) if stop_token else None]
# Remove all text after 3 newlines
text = text[: text.find(new_lines) if new_lines else None]
print("\n{}: {}".format(i, text))
print("\n" + 100 * '-')
```
|
Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition | a1e1c0fb14af7e4c4c2dfb2f35860da8b744f1b0 | 2022-07-24T10:47:37.000Z | [
"pytorch",
"wav2vec2",
"ru",
"dataset:Aniemore/resd",
"transformers",
"audio-classification",
"audio",
"emotion",
"emotion-recognition",
"emotion-classification",
"speech",
"license:mit",
"model-index"
] | audio-classification | false | Aniemore | null | Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition | 946 | 2 | transformers | 1,802 | ---
language: ru
tags:
- audio-classification
- audio
- emotion
- emotion-recognition
- emotion-classification
- speech
license: mit
datasets:
- Aniemore/resd
model-index:
- name: XLS-R Wav2Vec2 For Russian Speech Emotion Classification by Nikita Davidchuk
results:
- task:
name: Audio Emotion Recognition
type: audio-emotion-recognition
dataset:
name: Russian Emotional Speech Dialogs
type: Aniemore/resd
args: ru
metrics:
- name: accuracy
type: accuracy
value: 72%
---
# Prepare and importing
```python
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchaudio
from transformers import AutoConfig, AutoModel, Wav2Vec2FeatureExtractor
import librosa
import numpy as np
def speech_file_to_array_fn(path, sampling_rate):
speech_array, _sampling_rate = torchaudio.load(path)
resampler = torchaudio.transforms.Resample(_sampling_rate)
speech = resampler(speech_array).squeeze().numpy()
return speech
def predict(path, sampling_rate):
speech = speech_file_to_array_fn(path, sampling_rate)
inputs = feature_extractor(speech, sampling_rate=sampling_rate, return_tensors="pt", padding=True)
inputs = {key: inputs[key].to(device) for key in inputs}
with torch.no_grad():
logits = model_(**inputs).logits
scores = F.softmax(logits, dim=1).detach().cpu().numpy()[0]
outputs = [{"Emotion": config.id2label[i], "Score": f"{round(score * 100, 3):.1f}%"} for i, score in enumerate(scores)]
return outputs
```
# Evoking:
```python
TRUST = True
config = AutoConfig.from_pretrained('Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition', trust_remote_code=TRUST)
model_ = AutoModel.from_pretrained("Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition", trust_remote_code=TRUST)
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained("Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model_.to(device)
```
# Use case
```python
result = predict("/path/to/russian_audio_speech.wav", 16000)
print(result)
```
```python
# outputs
[{'Emotion': 'anger', 'Score': '0.0%'},
{'Emotion': 'disgust', 'Score': '100.0%'},
{'Emotion': 'enthusiasm', 'Score': '0.0%'},
{'Emotion': 'fear', 'Score': '0.0%'},
{'Emotion': 'happiness', 'Score': '0.0%'},
{'Emotion': 'neutral', 'Score': '0.0%'},
{'Emotion': 'sadness', 'Score': '0.0%'}]
```
# Results
| | precision | recall | f1-score | support |
|--------------|-----------|--------|----------|---------|
| anger | 0.97 | 0.86 | 0.92 | 44 |
| disgust | 0.71 | 0.78 | 0.74 | 37 |
| enthusiasm | 0.51 | 0.80 | 0.62 | 40 |
| fear | 0.80 | 0.62 | 0.70 | 45 |
| happiness | 0.66 | 0.70 | 0.68 | 44 |
| neutral | 0.81 | 0.66 | 0.72 | 38 |
| sadness | 0.79 | 0.59 | 0.68 | 32 |
| accuracy | | | 0.72 | 280 |
| macro avg | 0.75 | 0.72 | 0.72 | 280 |
| weighted avg | 0.75 | 0.72 | 0.73 | 280 |
# Citations
```
@misc{Aniemore,
author = {Артем Аментес, Илья Лубенец, Никита Давидчук},
title = {Открытая библиотека искусственного интеллекта для анализа и выявления эмоциональных оттенков речи человека},
year = {2022},
publisher = {Hugging Face},
journal = {Hugging Face Hub},
howpublished = {\url{https://huggingface.com/aniemore/Aniemore}},
email = {[email protected]}
}
``` |
DataikuNLP/tiny-random-bert | 7fb1dc6c16498e2028cc74afcf64319302fab101 | 2021-11-19T16:27:38.000Z | [
"pytorch",
"tf",
"bert",
"transformers"
] | null | false | DataikuNLP | null | DataikuNLP/tiny-random-bert | 944 | null | transformers | 1,803 | Entry not found |
mrm8488/mbart-large-finetuned-opus-en-es-translation | a9bfab17eba27c9bf319056017d610141ce137f0 | 2021-01-26T12:24:37.000Z | [
"pytorch",
"mbart",
"text2text-generation",
"en",
"es",
"dataset:opus100",
"transformers",
"translation",
"autotrain_compatible"
] | translation | false | mrm8488 | null | mrm8488/mbart-large-finetuned-opus-en-es-translation | 944 | 2 | transformers | 1,804 | ---
tags:
- translation
language:
- en
- es
datasets:
- opus100
---
### mbart-large-en-es
This is mbart-large-cc25, finetuned on opus100 for English to Spanish translation.
It scores BLEU **32.54** on test set.
|
sultan/BioM-ALBERT-xxlarge | c85649a7b3345b7de438e59c03f8f702e3aebc76 | 2021-10-12T21:23:31.000Z | [
"pytorch",
"albert",
"fill-mask",
"transformers",
"autotrain_compatible"
] | fill-mask | false | sultan | null | sultan/BioM-ALBERT-xxlarge | 944 | 1 | transformers | 1,805 | # BioM-Transformers: Building Large Biomedical Language Models with BERT, ALBERT and ELECTRA
# Abstract
The impact of design choices on the performance
of biomedical language models recently
has been a subject for investigation. In
this paper, we empirically study biomedical
domain adaptation with large transformer models
using different design choices. We evaluate
the performance of our pretrained models
against other existing biomedical language
models in the literature. Our results show that
we achieve state-of-the-art results on several
biomedical domain tasks despite using similar
or less computational cost compared to other
models in the literature. Our findings highlight
the significant effect of design choices on
improving the performance of biomedical language
models.
# Model Description
This model was pre-trained on PubMed Abstracts only with biomedical domain vocabulary for 264K steps with a batch size of 8192 on TPUv3-512 unit. In order to help researchers with limited resources to fine-tune larger models, we created an example with PyTorch XLA. PyTorch XLA (https://github.com/pytorch/xla) is a library that allows you to use PyTorch on TPU units, which is provided for free by Google Colab and Kaggle. Follow this example to work with PyTorch/XLA [Link](https://github.com/salrowili/BioM-Transformers/blob/main/examples/Fine_Tuning_Biomedical_Models_on_Text_Classification_Task_With_HuggingFace_Transformers_and_PyTorch_XLA.ipynb)
Check our GitHub repo at https://github.com/salrowili/BioM-Transformers for TensorFlow and GluonNLP checkpoints. We also updated this repo with a couple of examples on how to fine-tune LMs on text classification and questions answering tasks such as ChemProt, SQuAD, and BioASQ.
# Colab Notebook Examples
BioM-ELECTRA-LARGE on NER and ChemProt Task [![Open In Colab][COLAB]](https://colab.research.google.com/github/salrowili/BioM-Transformers/blob/main/examples/Example_of_NER_and_ChemProt_Task_on_TPU.ipynb)
BioM-ELECTRA-Large on SQuAD2.0 and BioASQ7B Factoid tasks [![Open In Colab][COLAB]](https://colab.research.google.com/github/salrowili/BioM-Transformers/blob/main/examples/Example_of_SQuAD2_0_and_BioASQ7B_tasks_with_BioM_ELECTRA_Large_on_TPU.ipynb)
BioM-ALBERT-xxlarge on SQuAD2.0 and BioASQ7B Factoid tasks [![Open In Colab][COLAB]](https://colab.research.google.com/github/salrowili/BioM-Transformers/blob/main/examples/Example_of_SQuAD2_0_and_BioASQ7B_tasks_with_BioM_ALBERT_xxlarge_on_TPU.ipynb)
Text Classification Task With HuggingFace Transformers and PyTorchXLA on Free TPU [![Open In Colab][COLAB]](https://colab.research.google.com/github/salrowili/BioM-Transformers/blob/main/examples/Fine_Tuning_Biomedical_Models_on_Text_Classification_Task_With_HuggingFace_Transformers_and_PyTorch_XLA.ipynb)
[COLAB]: https://colab.research.google.com/assets/colab-badge.svg
# Acknowledgment
We would like to acknowledge the support we have from Tensorflow Research Cloud (TFRC) team to grant us access to TPUv3 units.
# Citation
```bibtex
@inproceedings{alrowili-shanker-2021-biom,
title = "{B}io{M}-Transformers: Building Large Biomedical Language Models with {BERT}, {ALBERT} and {ELECTRA}",
author = "Alrowili, Sultan and
Shanker, Vijay",
booktitle = "Proceedings of the 20th Workshop on Biomedical Language Processing",
month = jun,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2021.bionlp-1.24",
pages = "221--227",
abstract = "The impact of design choices on the performance of biomedical language models recently has been a subject for investigation. In this paper, we empirically study biomedical domain adaptation with large transformer models using different design choices. We evaluate the performance of our pretrained models against other existing biomedical language models in the literature. Our results show that we achieve state-of-the-art results on several biomedical domain tasks despite using similar or less computational cost compared to other models in the literature. Our findings highlight the significant effect of design choices on improving the performance of biomedical language models.",
}
``` |
funnel-transformer/small-base | 373ecb760257d0059f1efdf58b3796d4d616ad0a | 2020-12-11T21:40:41.000Z | [
"pytorch",
"tf",
"funnel",
"feature-extraction",
"en",
"dataset:bookcorpus",
"dataset:wikipedia",
"dataset:gigaword",
"arxiv:2006.03236",
"transformers",
"license:apache-2.0"
] | feature-extraction | false | funnel-transformer | null | funnel-transformer/small-base | 943 | null | transformers | 1,806 | ---
language: en
license: apache-2.0
datasets:
- bookcorpus
- wikipedia
- gigaword
---
# Funnel Transformer small model (B4-4-4 without decoder)
Pretrained model on English language using a similar objective objective as [ELECTRA](https://huggingface.co/transformers/model_doc/electra.html). It was introduced in
[this paper](https://arxiv.org/pdf/2006.03236.pdf) and first released in
[this repository](https://github.com/laiguokun/Funnel-Transformer). This model is uncased: it does not make a difference
between english and English.
Disclaimer: The team releasing Funnel Transformer did not write a model card for this model so this model card has been
written by the Hugging Face team.
## Model description
Funnel Transformer is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it
was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of
publicly available data) with an automatic process to generate inputs and labels from those texts.
More precisely, a small language model corrupts the input texts and serves as a generator of inputs for this model, and
the pretraining objective is to predict which token is an original and which one has been replaced, a bit like a GAN training.
This way, the model learns an inner representation of the English language that can then be used to extract features
useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard
classifier using the features produced by the BERT model as inputs.
**Note:** This model does not contain the decoder, so it ouputs hidden states that have a sequence length of one fourth
of the inputs. It's good to use for tasks requiring a summary of the sentence (like sentence classification) but not if
you need one input per initial token. You should use the `small` model in that case.
## Intended uses & limitations
You can use the raw model to extract a vector representation of a given text, but it's mostly intended to
be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?filter=funnel-transformer) to look for
fine-tuned versions on a task that interests you.
Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
generation you should look at model like GPT2.
### How to use
Here is how to use this model to get the features of a given text in PyTorch:
```python
from transformers import FunnelTokenizer, FunnelBaseModel
tokenizer = FunnelTokenizer.from_pretrained("funnel-transformer/small-base")
model = FunnelBaseModel.from_pretrained("funnel-transformer/small-base")
text = "Replace me by any text you'd like."
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
```
and in TensorFlow:
```python
from transformers import FunnelTokenizer, TFFunnelBaseModel
tokenizer = FunnelTokenizer.from_pretrained("funnel-transformer/small-base")
model = TFFunnelBaseModel.from_pretrained("funnel-transformer/small-base")
text = "Replace me by any text you'd like."
encoded_input = tokenizer(text, return_tensors='tf')
output = model(encoded_input)
```
## Training data
The BERT model was pretrained on:
- [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of 11,038 unpublished books,
- [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia) (excluding lists, tables and headers),
- [Clue Web](https://lemurproject.org/clueweb12/), a dataset of 733,019,372 English web pages,
- [GigaWord](https://catalog.ldc.upenn.edu/LDC2011T07), an archive of newswire text data,
- [Common Crawl](https://commoncrawl.org/), a dataset of raw web pages.
### BibTeX entry and citation info
```bibtex
@misc{dai2020funneltransformer,
title={Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing},
author={Zihang Dai and Guokun Lai and Yiming Yang and Quoc V. Le},
year={2020},
eprint={2006.03236},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
```
|
codeparrot/codeparrot | 065248a99f051da363b1c2cbf05da943c8b6211b | 2022-06-24T08:28:28.000Z | [
"pytorch",
"tensorboard",
"gpt2",
"text-generation",
"code",
"dataset:codeparrot/codeparrot-clean-train",
"transformers",
"generation",
"model-index"
] | text-generation | false | codeparrot | null | codeparrot/codeparrot | 943 | 24 | transformers | 1,807 | ---
language: code
tags:
- code
- gpt2
- generation
datasets:
- codeparrot/codeparrot-clean-train
widget:
- text: "from transformer import"
example_title: "Transformers"
- text: "def print_hello_world():\n\t"
example_title: "Hello World!"
- text: "def get_file_size(filepath):"
example_title: "File size"
- text: "import numpy as"
example_title: "Numpy"
model-index:
- name: codeparrot
results:
- task:
name: Code Generation
type: code-generation
dataset:
name: "HumanEval"
type: openai_humaneval
metrics:
- name: pass@1
type: code_eval
value: 3.99
- name: pass@10
type: code_eval
value: 8.69
- name: pass@100
type: code_eval
value: 17.88
---
# CodeParrot 🦜
CodeParrot 🦜 is a GPT-2 model (1.5B parameters) trained to generate Python code. After the initial training and release of v1.0 we trained the model some more and released v1.1 (see below for details).
## Usage
You can load the CodeParrot model and tokenizer directly in `transformers`:
```Python
from transformers import AutoTokenizer, AutoModelWithLMHead
tokenizer = AutoTokenizer.from_pretrained("codeparrot/codeparrot")
model = AutoModelWithLMHead.from_pretrained("codeparrot/codeparrot")
inputs = tokenizer("def hello_world():", return_tensors="pt")
outputs = model(**inputs)
```
or with a `pipeline`:
```Python
from transformers import pipeline
pipe = pipeline("text-generation", model="codeparrot/codeparrot")
outputs = pipe("def hello_world():")
```
## Training
The model was trained on the cleaned [CodeParrot 🦜 dataset](https://huggingface.co/datasets/codeparrot/codeparrot-clean) in two steps. After the initial training (v1.0) the model was trained for another 30k steps resulting in v1.1 and you find the settings in the following table:
|Config| v1.0| v1.1|
|------|------------------|--------------------|
|Batch size| 512 | 512 |
|Context size| 1024 | 1024 |
|Training steps| 50'000| 30'000
|Gradient accumulation| 16| 16 |
|Gradient checkpointing| True| True |
|Learning rate| 2e-4 | 5e-5 |
|Weight decay | 0.1 | 0.1 |
|Warmup steps| 750 | 750 |
|Schedule| Cosine | Cosine |
The training was executed on 16 x A100 (40GB) GPUs. This setting amounts to roughly 26 + 15 billion tokens.
## Performance
We evaluated the model on OpenAI's [HumanEval](https://huggingface.co/datasets/openai_humaneval) benchmark which consists of programming challenges:
| Metric | v1.0 | v1.1 |
|--------|-----|-----|
|pass@1 | 3.58% | 3.99% |
|pass@10 | 8.03% | 8.69% |
|pass@100 | 14.96% | 17.88% |
The [pass@k metric](https://huggingface.co/metrics/code_eval) tells the probability that at least one out of k generations passes the tests.
## Resources
- Dataset: [full](https://huggingface.co/datasets/codeparrot/codeparrot-clean), [train](https://huggingface.co/datasets/codeparrot/codeparrot-clean-train), [valid](https://huggingface.co/datasets/codeparrot/codeparrot-clean-valid)
- Code: [repository](https://github.com/huggingface/transformers/tree/master/examples/research_projects/codeparrot)
- Spaces: [generation](), [highlighting]() |
Helsinki-NLP/opus-mt-it-fr | e2b19cab7c3ec41f1f314afc78a4e9423605f553 | 2021-09-10T13:53:00.000Z | [
"pytorch",
"marian",
"text2text-generation",
"it",
"fr",
"transformers",
"translation",
"license:apache-2.0",
"autotrain_compatible"
] | translation | false | Helsinki-NLP | null | Helsinki-NLP/opus-mt-it-fr | 942 | null | transformers | 1,808 | ---
tags:
- translation
license: apache-2.0
---
### opus-mt-it-fr
* source languages: it
* target languages: fr
* OPUS readme: [it-fr](https://github.com/Helsinki-NLP/OPUS-MT-train/blob/master/models/it-fr/README.md)
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download original weights: [opus-2020-01-16.zip](https://object.pouta.csc.fi/OPUS-MT-models/it-fr/opus-2020-01-16.zip)
* test set translations: [opus-2020-01-16.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/it-fr/opus-2020-01-16.test.txt)
* test set scores: [opus-2020-01-16.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/it-fr/opus-2020-01-16.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.it.fr | 67.9 | 0.792 |
|
UrukHan/t5-russian-spell | 9e1e7dda2468aac53b8e4aa1ae30dc8fd9e9c47c | 2022-04-04T18:55:49.000Z | [
"pytorch",
"tensorboard",
"t5",
"text2text-generation",
"transformers",
"generated_from_trainer",
"model-index",
"autotrain_compatible"
] | text2text-generation | false | UrukHan | null | UrukHan/t5-russian-spell | 942 | 1 | transformers | 1,809 | ---
tags:
- generated_from_trainer
model-index:
- name: t5-russian-spell
results: []
widget:
- text: "ывсем привет выныканалетоп армии и это двадцать пятый день спец операций на украине ет самый главной новости российские военные ракетами кинжалы калибр уничтожили крупную военную топливную базу украины ракетным ударом по населенному пункту под жетамиром уничтжены более стаукраинских военных в две тысячи двадцать втором году"
---
<!-- 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. -->
---
# t5-russian-spell
---
модель для исправление текста из распознаного аудио. моя модлеь для распознования аудио https://huggingface.co/UrukHan/wav2vec2-russian и его результаты можно закидывать в эту модель. тестил на видео случайном с ютюба
<table border="0">
<tr>
<td><b style="font-size:30px">Output wav2vec2</b></td>
<td><b style="font-size:30px">Output spell correcor</b></td>
</tr>
<tr>
<td>ывсем привет выныканалетоп армии и это двадцать пятый день спец операций на украине ет самый главной новости российские военные ракетами кинжалы калибр уничтожили крупную военную топливную базу украины ракетным ударом по населенному пункту под жетамиром уничтжены более стаукраинских военных в две тысячи двадцать втором году</td>
<td>Всем привет! Вы в курсе новостей от армии. И это 25 день спецопераций на Украине. Есть самые главные новости. Российские военные ракетами «Кинжалы» и «Кинжалы» калибра уничтожили крупную военную топливную базу Украины. Ракетным ударом по населенному пункту под Жетамиром уничтожены более ста украинских военных в 2022г.</td>
</tr>
</table>
#
---
Датасеты для обучения:
UrukHan/t5-russian-spell_I : https://huggingface.co/datasets/UrukHan/t5-russian-spell_I
UrukHan/t5-russian-spell_II : https://huggingface.co/datasets/UrukHan/t5-russian-spell_II
UrukHan/t5-russian-spell_III : https://huggingface.co/datasets/UrukHan/t5-russian-spell_III
---
# Запуск на вывод результатов пример работы с комментариями в колабе https://colab.research.google.com/drive/1ame2va9_NflYqy4RZ07HYmQ0moJYy7w2?usp=sharing :
#
```python
# Установим библиотеку трансформеров
!pip install transformers
# Импортируем библиотеки
from transformers import AutoModelForSeq2SeqLM, T5TokenizerFast
# Зададим название выбронной модели из хаба
MODEL_NAME = 'UrukHan/t5-russian-spell'
MAX_INPUT = 256
# Загрузка модели и токенизатора
tokenizer = T5TokenizerFast.from_pretrained(MODEL_NAME)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
# Входные данные (можно массив фраз или текст)
input_sequences = ['сеглдыя хорош ден', 'когд а вы прдет к нам в госи'] # или можно использовать одиночные фразы: input_sequences = 'сеглдыя хорош ден'
task_prefix = "Spell correct: " # Токенизирование данных
if type(input_sequences) != list: input_sequences = [input_sequences]
encoded = tokenizer(
[task_prefix + sequence for sequence in input_sequences],
padding="longest",
max_length=MAX_INPUT,
truncation=True,
return_tensors="pt",
)
predicts = model.generate(encoded) # # Прогнозирование
tokenizer.batch_decode(predicts, skip_special_tokens=True) # Декодируем данные
```
#
---
#Настроенный блокнот для запуска обучения и сохранения модели в свой репозиторий на huggingface hub:
#https://colab.research.google.com/drive/1H4IoasDqa2TEjGivVDp-4Pdpm0oxrCWd?usp=sharing
#
```python
# Установка библиотек
!pip install datasets
!apt install git-lfs
!pip install transformers
!pip install sentencepiece
!pip install rouge_score
# Импорт библиотек
import numpy as np
from datasets import Dataset
import tensorflow as
import nltk
from transformers import T5TokenizerFast, Seq2SeqTrainingArguments, Seq2SeqTrainer, AutoModelForSeq2SeqLM, DataCollatorForSeq2Seq
import torch
from transformers.optimization import Adafactor, AdafactorSchedule
from datasets import load_dataset, load_metric
# загрузка параметров
raw_datasets = load_dataset("xsum")
metric = load_metric("rouge")
nltk.download('punkt')
# Ввести свой ключ huggingface hyb
from huggingface_hub import notebook_login
notebook_login()
# Определение параметров
REPO = "t5-russian-spell" # Введите наазвание название репозитория
MODEL_NAME = "UrukHan/t5-russian-spell" # Введите наазвание выбранной модели из хаба
MAX_INPUT = 256 # Введите максимальную длинну входных данных в токенах (длинна входных фраз в словах (можно считать полслова токен))
MAX_OUTPUT = 256 # Введите максимальную длинну прогнозов в токенах (можно уменьшить для задач суммризации или других задач где выход короче)
BATCH_SIZE = 8
DATASET = 'UrukHan/t5-russian-spell_I' # Введите наазвание название датасета
# Загрузка датасета использование других типов данных опишу ниже
data = load_dataset(DATASET)
# Загрузка модели и токенизатора
tokenizer = T5TokenizerFast.from_pretrained(MODEL_NAME)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
model.config.max_length = MAX_OUTPUT # по умолчанию 20, поэтому во всех моделях прогнозы обрезаются выходные последовательности
# Закоментить после первого соъранения в репозиторий свой необъязательно
tokenizer.push_to_hub(repo_name)
train = data['train']
test = data['test'].train_test_split(0.02)['test'] # Уменьшил так тестовыу. выборку чтоб не ждать долго расчет ошибок между эпохами
data_collator = DataCollatorForSeq2Seq(tokenizer, model=model) #return_tensors="tf"
def compute_metrics(eval_pred):
predictions, labels = eval_pred
decoded_preds = tokenizer.batch_decode(predictions, skip_special_tokens=True)
# Replace -100 in the labels as we can't decode them.
labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True)
# Rouge expects a newline after each sentence
decoded_preds = ["\n".join(nltk.sent_tokenize(pred.strip())) for pred in decoded_preds]
decoded_labels = ["\n".join(nltk.sent_tokenize(label.strip())) for label in decoded_labels]
result = metric.compute(predictions=decoded_preds, references=decoded_labels, use_stemmer=True)
# Extract a few results
result = {key: value.mid.fmeasure * 100 for key, value in result.items()}
# Add mean generated length
prediction_lens = [np.count_nonzero(pred != tokenizer.pad_token_id) for pred in predictions]
result["gen_len"] = np.mean(prediction_lens)
return {k: round(v, 4) for k, v in result.items()}
training_args = Seq2SeqTrainingArguments(
output_dir = REPO,
#overwrite_output_dir=True,
evaluation_strategy='steps',
#learning_rate=2e-5,
eval_steps=5000,
save_steps=5000,
num_train_epochs=1,
predict_with_generate=True,
per_device_train_batch_size=BATCH_SIZE,
per_device_eval_batch_size=BATCH_SIZE,
fp16=True,
save_total_limit=2,
#generation_max_length=256,
#generation_num_beams=4,
weight_decay=0.005,
#logging_dir='logs',
push_to_hub=True,
)
# Выберем вручную оптимизатор. Т5 в оригинальной архитектуре использует Адафактор оптимизатор
optimizer = Adafactor(
model.parameters(),
lr=1e-5,
eps=(1e-30, 1e-3),
clip_threshold=1.0,
decay_rate=-0.8,
beta1=None,
weight_decay=0.0,
relative_step=False,
scale_parameter=False,
warmup_init=False,
)
lr_scheduler = AdafactorSchedule(optimizer)
trainer = Seq2SeqTrainer(
model=model,
args=training_args,
train_dataset = train,
eval_dataset = test,
optimizers = (optimizer, lr_scheduler),
tokenizer = tokenizer,
compute_metrics=compute_metrics
)
trainer.train()
trainer.push_to_hub()
```
#
---
# Пример конвертации массивов для данной сети
#
```python
input_data = ['удач почти отнее отвернулась', 'в хааоде проведения чемпиониавта мира дветысячивосемнандцтая лгодаа']
output_data = ['Удача почти от нее отвернулась', 'в ходе проведения чемпионата мира две тысячи восемнадцатого года']
# Токенизируем входные данные
task_prefix = "Spell correct: "
input_sequences = input_data
encoding = tokenizer(
[task_prefix + sequence for sequence in input_sequences],
padding="longest",
max_length=MAX_INPUT,
truncation=True,
return_tensors="pt",
)
input_ids, attention_mask = encoding.input_ids, encoding.attention_mask
# Токенизируем выходные данные
target_encoding = tokenizer(output_data, padding="longest", max_length=MAX_OUTPUT, truncation=True)
labels = target_encoding.input_ids
# replace padding token id's of the labels by -100
labels = torch.tensor(labels)
labels[labels == tokenizer.pad_token_id] = -100'''
# Конвертируем наши данные в формат dataset
data = Dataset.from_pandas(pd.DataFrame({'input_ids': list(np.array(input_ids)), 'attention_mask': list(np.array(attention_mask)), 'labels': list(np.array(labels))}))
data = data.train_test_split(0.02)
# и получим на вход сети для нашешго trainer: train_dataset = data['train'], eval_dataset = data['test'] |
Stancld/longt5-tglobal-large-16384-pubmed-3k_steps | 6949726515747477615ee1cecb31ff5d34ca3add | 2022-06-20T15:45:47.000Z | [
"pytorch",
"jax",
"longt5",
"text2text-generation",
"en",
"dataset:ccdv/pubmed-summarization",
"arxiv:2112.07916",
"arxiv:1910.10683",
"transformers",
"license:apache-2.0",
"autotrain_compatible"
] | text2text-generation | false | Stancld | null | Stancld/longt5-tglobal-large-16384-pubmed-3k_steps | 941 | 7 | transformers | 1,810 | ---
language: en
datasets:
- ccdv/pubmed-summarization
license: apache-2.0
---
## Introduction
[Google's LongT5: Efficient Text-To-Text Transformer for Long Sequences](https://arxiv.org/pdf/2112.07916.pdf) introduced as an extension of a successful [T5 model](https://arxiv.org/pdf/1910.10683.pdf).
This is an unofficial *longt5-large-16384-pubmed-3k_steps* checkpoint. I.e., this is a large configuration of the LongT5 model with a `transient-global` attention fine-tuned on [pubmed summarization dataset](https://huggingface.co/datasets/ccdv/pubmed-summarization) for 3,000 training steps. It may be worth continuing in the fine-tuning as we did not train the model until the convergence.
## Results and Fine-tuning Details
The fine-tuned model achieves the following results on the evaluation set using `beam_search=3` and without any specific calibration of generation parameters are presented below, altogether with the results from the original paper (the original scores are higher, very likely due to a higher number of training steps).
| Metric | Score | Score (original paper)
| --- | --- | --- |
| Rouge-1 | 47.44 | 49.98 |
| Rouge-2 | 22.68 | 24.69 |
| Rouge-L | 29.83 | x |
| Rouge-Lsum | 43.13 | 46.46 |
The training parameters follow the ones specified in the paper. We accumulated batch size to 128 examples and used `Adafactor` optimizer with a constant learning rate `0.001`. The full training hyper-parameters and logs can be found via the following [W&B run](https://wandb.ai/stancld/LongT5/runs/1lwncl8a?workspace=user-stancld). The model was trained using the [HuggingFace's trainer](https://github.com/huggingface/transformers/blob/main/src/transformers/trainer_seq2seq.py).
The only specific adjustment, I made for the training, was dropping very short input articles (less than 16 words (a bit of mistake, should be less than 16 tokens)) as this sequences do not contribute to gradient creation in the *transient-global* attention, which resulted in training crashes when DDP used.
## Usage
```python
LONG_ARTICLE = """"anxiety affects quality of life in those living
with parkinson 's disease ( pd ) more so than
overall cognitive status , motor deficits , apathy
, and depression [ 13 ] . although anxiety and
depression are often related and coexist in pd
patients , recent research suggests that anxiety
rather than depression is the most prominent and
prevalent mood disorder in pd [ 5 , 6 ] . yet ,
our current understanding of anxiety and its
impact on cognition in pd , as well as its neural
basis and best treatment practices , remains
meager and lags far behind that of depression .
overall , neuropsychiatric symptoms in pd have
been shown to be negatively associated with
cognitive performance . for example , higher
depression scores have been correlated with lower
scores on the mini - mental state exam ( mmse ) [
8 , 9 ] as well as tests of memory and executive
functions ( e.g. , attention ) [ 1014 ] . likewise
, apathy and anhedonia in pd patients have been
associated with executive dysfunction [ 10 , 1523
] . however , few studies have specifically
investigated the relationship between anxiety and
cognition in pd . one study showed a strong
negative relationship between anxiety ( both state
and trait ) and overall cognitive performance (
measured by the total of the repeatable battery
for the assessment of neuropsychological status
index ) within a sample of 27 pd patients .
furthermore , trait anxiety was negatively
associated with each of the cognitive domains
assessed by the rbans ( i.e. , immediate memory ,
visuospatial construction , language , attention ,
and delayed memory ) . two further studies have
examined whether anxiety differentially affects
cognition in patients with left - sided dominant
pd ( lpd ) versus right - sided dominant pd ( rpd
) ; however , their findings were inconsistent .
the first study found that working memory
performance was worse in lpd patients with anxiety
compared to rpd patients with anxiety , whereas
the second study reported that , in lpd , apathy
but not anxiety was associated with performance on
nonverbally mediated executive functions and
visuospatial tasks ( e.g. , tmt - b , wms - iii
spatial span ) , while in rpd , anxiety but not
apathy significantly correlated with performance
on verbally mediated tasks ( e.g. , clock reading
test and boston naming test ) . furthermore ,
anxiety was significantly correlated with
neuropsychological measures of attention and
executive and visuospatial functions . taken
together , it is evident that there are limited
and inconsistent findings describing the
relationship between anxiety and cognition in pd
and more specifically how anxiety might influence
particular domains of cognition such as attention
and memory and executive functioning . it is also
striking that , to date , no study has examined
the influence of anxiety on cognition in pd by
directly comparing groups of pd patients with and
without anxiety while excluding depression . given
that research on healthy young adults suggests
that anxiety reduces processing capacity and
impairs processing efficiency , especially in the
central executive and attentional systems of
working memory [ 26 , 27 ] , we hypothesized that
pd patients with anxiety would show impairments in
attentional set - shifting and working memory
compared to pd patients without anxiety .
furthermore , since previous work , albeit limited
, has focused on the influence of symptom
laterality on anxiety and cognition , we also
explored this relationship . seventeen pd patients
with anxiety and thirty - three pd patients
without anxiety were included in this study ( see
table 1 ) . the cross - sectional data from these
participants was taken from a patient database
that has been compiled over the past 8 years (
since 2008 ) at the parkinson 's disease research
clinic at the brain and mind centre , university
of sydney . inclusion criteria involved a
diagnosis of idiopathic pd according to the united
kingdom parkinson 's disease society brain bank
criteria and were confirmed by a neurologist (
sjgl ) . patients also had to have an adequate
proficiency in english and have completed a full
neuropsychological assessment . ten patients in
this study ( 5 pd with anxiety ; 5 pd without
anxiety ) were taking psychotropic drugs ( i.e. ,
benzodiazepine or selective serotonin reuptake
inhibitor ) . patients were also excluded if they
had other neurological disorders , psychiatric
disorders other than affective disorders ( such as
anxiety ) , or if they reported a score greater
than six on the depression subscale of the
hospital anxiety and depression scale ( hads ) .
thus , all participants who scored within a
depressed ( hads - d > 6 ) range were excluded
from this study , in attempt to examine a refined
sample of pd patients with and without anxiety in
order to determine the independent effect of
anxiety on cognition . this research was approved
by the human research ethics committee of the
university of sydney , and written informed
consent was obtained from all participants . self
- reported hads was used to assess anxiety in pd
and has been previously shown to be a useful
measure of clinical anxiety in pd . a cut - off
score of > 8 on the anxiety subscale of the hads (
hads - a ) was used to identify pd cases with
anxiety ( pda+ ) , while a cut - off score of < 6
on the hads - a was used to identify pd cases
without anxiety ( pda ) . this criterion was more
stringent than usual ( > 7 cut - off score ) , in
effort to create distinct patient groups . the
neurological evaluation rated participants
according to hoehn and yahr ( h&y ) stages and
assessed their motor symptoms using part iii of
the revised mds task force unified parkinson 's
disease rating scale ( updrs ) . in a similar way
this was determined by calculating a total left
and right score from rigidity items 3035 ,
voluntary movement items 3643 , and tremor items
5057 from the mds - updrs part iii ( see table 1 )
. processing speed was assessed using the trail
making test , part a ( tmt - a , z - score ) .
attentional set - shifting was measured using the
trail making test , part b ( tmt - b , z - score )
. working memory was assessed using the digit span
forward and backward subtest of the wechsler
memory scale - iii ( raw scores ) . language was
assessed with semantic and phonemic verbal fluency
via the controlled oral word associated test (
cowat animals and letters , z - score ) . the
ability to retain learned verbal memory was
assessed using the logical memory subtest from the
wechsler memory scale - iii ( lm - i z - score ,
lm - ii z - score , % lm retention z - score ) .
the mini - mental state examination ( mmse )
demographic , clinical , and neuropsychological
variables were compared between the two groups
with the independent t - test or mann whitney u
test , depending on whether the variable met
parametric assumptions . chi - square tests were
used to examine gender and symptom laterality
differences between groups . all analyses employed
an alpha level of p < 0.05 and were two - tailed .
spearman correlations were performed separately in
each group to examine associations between anxiety
and/or depression ratings and cognitive functions
. as expected , the pda+ group reported
significant greater levels of anxiety on the hads
- a ( u = 0 , p < 0.001 ) and higher total score
on the hads ( u = 1 , p < 0.001 ) compared to the
pda group ( table 1 ) . groups were matched in age
( t(48 ) = 1.31 , p = 0.20 ) , disease duration (
u = 259 , p = 0.66 ) , updrs - iii score ( u =
250.5 , p = 0.65 ) , h&y ( u = 245 , p = 0.43 ) ,
ledd ( u = 159.5 , p = 0.80 ) , and depression (
hads - d ) ( u = 190.5 , p = 0.06 ) . additionally
, all groups were matched in the distribution of
gender ( = 0.098 , p = 0.75 ) and side - affected
( = 0.765 , p = 0.38 ) . there were no group
differences for tmt - a performance ( u = 256 , p
= 0.62 ) ( table 2 ) ; however , the pda+ group
had worse performance on the trail making test
part b ( t(46 ) = 2.03 , p = 0.048 ) compared to
the pda group ( figure 1 ) . the pda+ group also
demonstrated significantly worse performance on
the digit span forward subtest ( t(48 ) = 2.22 , p
= 0.031 ) and backward subtest ( u = 190.5 , p =
0.016 ) compared to the pda group ( figures 2(a )
and 2(b ) ) . neither semantic verbal fluency (
t(47 ) = 0.70 , p = 0.49 ) nor phonemic verbal
fluency ( t(47 ) = 0.39 , p = 0.70 ) differed
between groups . logical memory i immediate recall
test ( u = 176 , p = 0.059 ) showed a trend that
the pda+ group had worse new verbal learning and
immediate recall abilities than the pda group .
however , logical memory ii test performance ( u =
219 , p = 0.204 ) and logical memory % retention (
u = 242.5 , p = 0.434 ) did not differ between
groups . there were also no differences between
groups in global cognition ( mmse ) ( u = 222.5 ,
p = 0.23 ) . participants were split into lpd and
rpd , and then further group differences were
examined between pda+ and pda. importantly , the
groups remained matched in age , disease duration
, updrs - iii , dde , h&y stage , and depression
but remained significantly different on self -
reported anxiety . lpda+ demonstrated worse
performance on the digit span forward test ( t(19
) = 2.29 , p = 0.033 ) compared to lpda , whereas
rpda+ demonstrated worse performance on the digit
span backward test ( u = 36.5 , p = 0.006 ) , lm -
i immediate recall ( u = 37.5 , p = 0.008 ) , and
lm - ii ( u = 45.0 , p = 0.021 ) but not lm %
retention ( u = 75.5 , p = 0.39 ) compared to
rpda. this study is the first to directly compare
cognition between pd patients with and without
anxiety . the findings confirmed our hypothesis
that anxiety negatively influences attentional set
- shifting and working memory in pd . more
specifically , we found that pd patients with
anxiety were more impaired on the trail making
test part b which assessed attentional set -
shifting , on both digit span tests which assessed
working memory and attention , and to a lesser
extent on the logical memory test which assessed
memory and new verbal learning compared to pd
patients without anxiety . taken together , these
findings suggest that anxiety in pd may reduce
processing capacity and impair processing
efficiency , especially in the central executive
and attentional systems of working memory in a
similar way as seen in young healthy adults [ 26 ,
27 ] . although the neurobiology of anxiety in pd
remains unknown , many researchers have postulated
that anxiety disorders are related to
neurochemical changes that occur during the early
, premotor stages of pd - related degeneration [
37 , 38 ] such as nigrostriatal dopamine depletion
, as well as cell loss within serotonergic and
noradrenergic brainstem nuclei ( i.e. , raphe
nuclei and locus coeruleus , resp . , which
provide massive inputs to corticolimbic regions )
. over time , chronic dysregulation of
adrenocortical and catecholamine functions can
lead to hippocampal damage as well as
dysfunctional prefrontal neural circuitries [ 39 ,
40 ] , which play a key role in memory and
attention . recent functional neuroimaging work
has suggested that enhanced hippocampal activation
during executive functioning and working memory
tasks may represent compensatory processes for
impaired frontostriatal functions in pd patients
compared to controls . therefore , chronic stress
from anxiety , for example , may disrupt
compensatory processes in pd patients and explain
the cognitive impairments specifically in working
memory and attention seen in pd patients with
anxiety . it has also been suggested that
hyperactivation within the putamen may reflect a
compensatory striatal mechanism to maintain normal
working memory performance in pd patients ;
however , losing this compensatory activation has
been shown to contribute to poor working memory
performance . anxiety in mild pd has been linked
to reduced putamen dopamine uptake which becomes
more extensive as the disease progresses . this
further supports the notion that anxiety may
disrupt compensatory striatal mechanisms as well ,
providing another possible explanation for the
cognitive impairments observed in pd patients with
anxiety in this study . noradrenergic and
serotonergic systems should also be considered
when trying to explain the mechanisms by which
anxiety may influence cognition in pd . although
these neurotransmitter systems are relatively
understudied in pd cognition , treating the
noradrenergic and serotonergic systems has shown
beneficial effects on cognition in pd . selective
serotonin reuptake inhibitor , citalopram , was
shown to improve response inhibition deficits in
pd , while noradrenaline reuptake blocker ,
atomoxetine , has been recently reported to have
promising effects on cognition in pd [ 45 , 46 ] .
overall , very few neuroimaging studies have been
conducted in pd in order to understand the neural
correlates of pd anxiety and its underlying neural
pathology . future research should focus on
relating anatomical changes and neurochemical
changes to neural activation in order to gain a
clearer understanding on how these pathologies
affect anxiety in pd . to further understand how
anxiety and cognitive dysfunction are related ,
future research should focus on using advanced
structural and function imaging techniques to
explain both cognitive and neural breakdowns that
are associated with anxiety in pd patients .
research has indicated that those with amnestic
mild cognitive impairment who have more
neuropsychiatric symptoms have a greater risk of
developing dementia compared to those with fewer
neuropsychiatric symptoms . future studies should
also examine whether treating neuropsychiatric
symptoms might impact the progression of cognitive
decline and improve cognitive impairments in pd
patients . previous studies have used pd symptom
laterality as a window to infer asymmetrical
dysfunction of neural circuits . for example , lpd
patients have greater inferred right hemisphere
pathology , whereas rpd patients have greater
inferred left hemisphere pathology . thus ,
cognitive domains predominantly subserved by the
left hemisphere ( e.g. , verbally mediated tasks
of executive function and verbal memory ) might be
hypothesized to be more affected in rpd than lpd ;
however , this remains controversial . it has also
been suggested that since anxiety is a common
feature of left hemisphere involvement [ 48 , 49 ]
, cognitive domains subserved by the left
hemisphere may also be more strongly related to
anxiety . results from this study showed selective
verbal memory deficits in rpd patients with
anxiety compared to rpd without anxiety , whereas
lpd patients with anxiety had greater attentional
/ working memory deficits compared to lpd without
anxiety . although these results align with
previous research , interpretations of these
findings should be made with caution due to the
small sample size in the lpd comparison
specifically . recent work has suggested that the
hads questionnaire may underestimate the burden of
anxiety related symptomology and therefore be a
less sensitive measure of anxiety in pd [ 30 , 50
] . in addition , our small sample size also
limited the statistical power for detecting
significant findings . based on these limitations
, our findings are likely conservative and
underrepresent the true impact anxiety has on
cognition in pd . additionally , the current study
employed a very brief neuropsychological
assessment including one or two tests for each
cognitive domain . future studies are encouraged
to collect a more complex and comprehensive
battery from a larger sample of pd participants in
order to better understand the role anxiety plays
on cognition in pd . another limitation of this
study was the absence of diagnostic interviews to
characterize participants ' psychiatric symptoms
and specify the type of anxiety disorders included
in this study . future studies should perform
diagnostic interviews with participants ( e.g. ,
using dsm - v criteria ) rather than relying on
self - reported measures to group participants ,
in order to better understand whether the type of
anxiety disorder ( e.g. , social anxiety , phobias
, panic disorders , and generalized anxiety )
influences cognitive performance differently in pd
. one advantage the hads questionnaire provided
over other anxiety scales was that it assessed
both anxiety and depression simultaneously and
allowed us to control for coexisting depression .
although there was a trend that the pda+ group
self - reported higher levels of depression than
the pda group , all participants included in the
study scored < 6 on the depression subscale of the
hads . controlling for depression while assessing
anxiety has been identified as a key shortcoming
in the majority of recent work . considering many
previous studies have investigated the influence
of depression on cognition in pd without
accounting for the presence of anxiety and the
inconsistent findings reported to date , we
recommend that future research should try to
disentangle the influence of anxiety versus
depression on cognitive impairments in pd .
considering the growing number of clinical trials
for treating depression , there are few if any for
the treatment of anxiety in pd . anxiety is a key
contributor to decreased quality of life in pd and
greatly requires better treatment options .
moreover , anxiety has been suggested to play a
key role in freezing of gait ( fog ) , which is
also related to attentional set - shifting [ 52 ,
53 ] . future research should examine the link
between anxiety , set - shifting , and fog , in
order to determine whether treating anxiety might
be a potential therapy for improving fog ."""
import torch
from transformers import AutoTokenizer, LongT5ForConditionalGeneration
tokenizer = AutoTokenizer.from_pretrained("Stancld/longt5-tglobal-large-16384-pubmed-3k_steps")
input_ids = tokenizer(LONG_ARTICLE, return_tensors="pt").input_ids.to("cuda")
model = LongT5ForConditionalGeneration.from_pretrained("Stancld/longt5-tglobal-large-16384-pubmed-3k_steps", return_dict_in_generate=True).to("cuda")
sequences = model.generate(input_ids).sequences
summary = tokenizer.batch_decode(sequences)
``` |
prithivida/passive_to_active_styletransfer | 81b7ef1c02f244ac9030d928f2c5a01a037c2430 | 2021-06-23T13:45:25.000Z | [
"pytorch",
"t5",
"text2text-generation",
"transformers",
"autotrain_compatible"
] | text2text-generation | false | prithivida | null | prithivida/passive_to_active_styletransfer | 940 | 3 | transformers | 1,811 | ## This model belongs to the Styleformer project
[Please refer to github page](https://github.com/PrithivirajDamodaran/Styleformer)
|
Helsinki-NLP/opus-mt-cy-en | 775c85089bc7a55c8203bff544e9fa34cd4ba7ca | 2021-09-09T21:29:44.000Z | [
"pytorch",
"marian",
"text2text-generation",
"cy",
"en",
"transformers",
"translation",
"license:apache-2.0",
"autotrain_compatible"
] | translation | false | Helsinki-NLP | null | Helsinki-NLP/opus-mt-cy-en | 938 | null | transformers | 1,812 | ---
tags:
- translation
license: apache-2.0
---
### opus-mt-cy-en
* source languages: cy
* target languages: en
* OPUS readme: [cy-en](https://github.com/Helsinki-NLP/OPUS-MT-train/blob/master/models/cy-en/README.md)
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download original weights: [opus-2019-12-18.zip](https://object.pouta.csc.fi/OPUS-MT-models/cy-en/opus-2019-12-18.zip)
* test set translations: [opus-2019-12-18.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/cy-en/opus-2019-12-18.test.txt)
* test set scores: [opus-2019-12-18.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/cy-en/opus-2019-12-18.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.cy.en | 33.0 | 0.525 |
|
nreimers/MiniLMv2-L6-H768-distilled-from-RoBERTa-Large | 242150e6be5fb4f7cdb66dbc6bffac542ff828ca | 2021-06-20T19:02:48.000Z | [
"pytorch",
"roberta",
"fill-mask",
"transformers",
"autotrain_compatible"
] | fill-mask | false | nreimers | null | nreimers/MiniLMv2-L6-H768-distilled-from-RoBERTa-Large | 935 | null | transformers | 1,813 | # MiniLMv2
This is a MiniLMv2 model from: [https://github.com/microsoft/unilm](https://github.com/microsoft/unilm/tree/master/minilm) |
MoritzLaurer/DeBERTa-v3-xsmall-mnli-fever-anli-ling-binary | e239423547914eda65121a26121768656f7a6400 | 2022-07-28T16:23:45.000Z | [
"pytorch",
"deberta-v2",
"text-classification",
"en",
"dataset:multi_nli",
"dataset:anli",
"dataset:fever",
"dataset:lingnli",
"arxiv:2104.07179",
"arxiv:2111.09543",
"transformers",
"zero-shot-classification",
"license:mit"
] | zero-shot-classification | false | MoritzLaurer | null | MoritzLaurer/DeBERTa-v3-xsmall-mnli-fever-anli-ling-binary | 934 | null | transformers | 1,814 | ---
language:
- en
license: mit
tags:
- text-classification
- zero-shot-classification
metrics:
- accuracy
datasets:
- multi_nli
- anli
- fever
- lingnli
pipeline_tag: zero-shot-classification
---
# DeBERTa-v3-xsmall-mnli-fever-anli-ling-binary
## Model description
This model was trained on 782 357 hypothesis-premise pairs from 4 NLI datasets: [MultiNLI](https://huggingface.co/datasets/multi_nli), [Fever-NLI](https://github.com/easonnie/combine-FEVER-NSMN/blob/master/other_resources/nli_fever.md), [LingNLI](https://arxiv.org/abs/2104.07179) and [ANLI](https://github.com/facebookresearch/anli).
Note that the model was trained on binary NLI to predict either "entailment" or "not-entailment". This is specifically designed for zero-shot classification, where the difference between "neutral" and "contradiction" is irrelevant.
The base model is [DeBERTa-v3-xsmall from Microsoft](https://huggingface.co/microsoft/deberta-v3-xsmall). The v3 variant of DeBERTa substantially outperforms previous versions of the model by including a different pre-training objective, see the [DeBERTa-V3 paper](https://arxiv.org/abs/2111.09543).
For highest performance (but less speed), I recommend using https://huggingface.co/MoritzLaurer/DeBERTa-v3-large-mnli-fever-anli-ling-wanli.
## Intended uses & limitations
#### How to use the model
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
model_name = "MoritzLaurer/DeBERTa-v3-xsmall-mnli-fever-anli-ling-binary"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
premise = "I first thought that I liked the movie, but upon second thought it was actually disappointing."
hypothesis = "The movie was good."
input = tokenizer(premise, hypothesis, truncation=True, return_tensors="pt")
output = model(input["input_ids"].to(device)) # device = "cuda:0" or "cpu"
prediction = torch.softmax(output["logits"][0], -1).tolist()
label_names = ["entailment", "not_entailment"]
prediction = {name: round(float(pred) * 100, 1) for pred, name in zip(prediction, label_names)}
print(prediction)
```
### Training data
This model was trained on 782 357 hypothesis-premise pairs from 4 NLI datasets: [MultiNLI](https://huggingface.co/datasets/multi_nli), [Fever-NLI](https://github.com/easonnie/combine-FEVER-NSMN/blob/master/other_resources/nli_fever.md), [LingNLI](https://arxiv.org/abs/2104.07179) and [ANLI](https://github.com/facebookresearch/anli).
### Training procedure
DeBERTa-v3-xsmall-mnli-fever-anli-ling-binary was trained using the Hugging Face trainer with the following hyperparameters.
```
training_args = TrainingArguments(
num_train_epochs=5, # total number of training epochs
learning_rate=2e-05,
per_device_train_batch_size=32, # batch size per device during training
per_device_eval_batch_size=32, # batch size for evaluation
warmup_ratio=0.1, # number of warmup steps for learning rate scheduler
weight_decay=0.06, # strength of weight decay
fp16=True # mixed precision training
)
```
### Eval results
The model was evaluated using the binary test sets for MultiNLI, ANLI, LingNLI and the binary dev set for Fever-NLI (two classes instead of three). The metric used is accuracy.
dataset | mnli-m-2c | mnli-mm-2c | fever-nli-2c | anli-all-2c | anli-r3-2c | lingnli-2c
--------|---------|----------|---------|----------|----------|------
accuracy | 0.925 | 0.922 | 0.892 | 0.676 | 0.665 | 0.888
speed (text/sec, CPU, 128 batch) | 6.0 | 6.3 | 3.0 | 5.8 | 5.0 | 7.6
speed (text/sec, GPU Tesla P100, 128 batch) | 473 | 487 | 230 | 390 | 340 | 586
## Limitations and bias
Please consult the original DeBERTa paper and literature on different NLI datasets for potential biases.
## Citation
If you use this model, please cite: Laurer, Moritz, Wouter van Atteveldt, Andreu Salleras Casas, and Kasper Welbers. 2022. ‘Less Annotating, More Classifying – Addressing the Data Scarcity Issue of Supervised Machine Learning with Deep Transfer Learning and BERT - NLI’. Preprint, June. Open Science Framework. https://osf.io/74b8k.
### Ideas for cooperation or questions?
If you have questions or ideas for cooperation, contact me at m{dot}laurer{at}vu{dot}nl or [LinkedIn](https://www.linkedin.com/in/moritz-laurer/)
### Debugging and issues
Note that DeBERTa-v3 was released on 06.12.21 and older versions of HF Transformers seem to have issues running the model (e.g. resulting in an issue with the tokenizer). Using Transformers>=4.13 might solve some issues. |
microsoft/swin-base-patch4-window12-384-in22k | c2c8cfc218cfcf3f43091c4476137ed4f2fed9a2 | 2022-05-16T18:01:06.000Z | [
"pytorch",
"tf",
"swin",
"image-classification",
"dataset:imagenet-21k",
"arxiv:2103.14030",
"transformers",
"vision",
"license:apache-2.0"
] | image-classification | false | microsoft | null | microsoft/swin-base-patch4-window12-384-in22k | 932 | null | transformers | 1,815 | ---
license: apache-2.0
tags:
- vision
- image-classification
datasets:
- imagenet-21k
widget:
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/tiger.jpg
example_title: Tiger
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg
example_title: Teapot
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/palace.jpg
example_title: Palace
---
# Swin Transformer (large-sized model)
Swin Transformer model pre-trained on ImageNet-21k (14 million images, 21,841 classes) at resolution 384x384. It was introduced in the paper [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) by Liu et al. and first released in [this repository](https://github.com/microsoft/Swin-Transformer).
Disclaimer: The team releasing Swin Transformer did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
The Swin Transformer is a type of Vision Transformer. It builds hierarchical feature maps by merging image patches (shown in gray) in deeper layers and has linear computation complexity to input image size due to computation of self-attention only within each local window (shown in red). It can thus serve as a general-purpose backbone for both image classification and dense recognition tasks. In contrast, previous vision Transformers produce feature maps of a single low resolution and have quadratic computation complexity to input image size due to computation of self-attention globally.
[Source](https://paperswithcode.com/method/swin-transformer)
## Intended uses & limitations
You can use the raw model for image classification. See the [model hub](https://huggingface.co/models?search=swin) to look for
fine-tuned versions on a task that interests you.
### How to use
Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 1,000 ImageNet classes:
```python
from transformers import AutoFeatureExtractor, SwinForImageClassification
from PIL import Image
import requests
url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)
feature_extractor = AutoFeatureExtractor.from_pretrained("microsoft/swin-base-patch4-window12-384-in22k")
model = SwinForImageClassification.from_pretrained("microsoft/swin-base-patch4-window12-384-in22k")
inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits
# model predicts one of the 1000 ImageNet classes
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])
```
For more code examples, we refer to the [documentation](https://huggingface.co/transformers/model_doc/swin.html#).
### BibTeX entry and citation info
```bibtex
@article{DBLP:journals/corr/abs-2103-14030,
author = {Ze Liu and
Yutong Lin and
Yue Cao and
Han Hu and
Yixuan Wei and
Zheng Zhang and
Stephen Lin and
Baining Guo},
title = {Swin Transformer: Hierarchical Vision Transformer using Shifted Windows},
journal = {CoRR},
volume = {abs/2103.14030},
year = {2021},
url = {https://arxiv.org/abs/2103.14030},
eprinttype = {arXiv},
eprint = {2103.14030},
timestamp = {Thu, 08 Apr 2021 07:53:26 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2103-14030.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
``` |
roberta-large-openai-detector | 5002d695ecf610d8bbfb1fa0d14f1575185b4915 | 2022-07-22T08:07:41.000Z | [
"pytorch",
"jax",
"roberta",
"text-classification",
"en",
"dataset:bookcorpus",
"dataset:wikipedia",
"arxiv:1904.09751",
"arxiv:1910.09700",
"transformers",
"exbert",
"license:mit"
] | text-classification | false | null | null | roberta-large-openai-detector | 928 | 1 | transformers | 1,816 | ---
language: en
license: mit
tags:
- exbert
datasets:
- bookcorpus
- wikipedia
---
# RoBERTa Large OpenAI Detector
## Table of Contents
- [Model Details](#model-details)
- [Uses](#uses)
- [Risks, Limitations and Biases](#risks-limitations-and-biases)
- [Training](#training)
- [Evaluation](#evaluation)
- [Environmental Impact](#environmental-impact)
- [Technical Specifications](#technical-specifications)
- [Citation Information](#citation-information)
- [Model Card Authors](#model-card-authors)
- [How To Get Started With the Model](#how-to-get-started-with-the-model)
## Model Details
**Model Description:** RoBERTa large OpenAI Detector is the GPT-2 output detector model, obtained by fine-tuning a RoBERTa large model with the outputs of the 1.5B-parameter GPT-2 model. The model can be used to predict if text was generated by a GPT-2 model. This model was released by OpenAI at the same time as OpenAI released the weights of the [largest GPT-2 model](https://huggingface.co/gpt2-xl), the 1.5B parameter version.
- **Developed by:** OpenAI, see [GitHub Repo](https://github.com/openai/gpt-2-output-dataset/tree/master/detector) and [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf) for full author list
- **Model Type:** Fine-tuned transformer-based language model
- **Language(s):** English
- **License:** MIT
- **Related Models:** [RoBERTa large](https://huggingface.co/roberta-large), [GPT-XL (1.5B parameter version)](https://huggingface.co/gpt2-xl), [GPT-Large (the 774M parameter version)](https://huggingface.co/gpt2-large), [GPT-Medium (the 355M parameter version)](https://huggingface.co/gpt2-medium) and [GPT-2 (the 124M parameter version)](https://huggingface.co/gpt2)
- **Resources for more information:**
- [Research Paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf) (see, in particular, the section beginning on page 12 about Automated ML-based detection).
- [GitHub Repo](https://github.com/openai/gpt-2-output-dataset/tree/master/detector)
- [OpenAI Blog Post](https://openai.com/blog/gpt-2-1-5b-release/)
- [Explore the detector model here](https://huggingface.co/openai-detector )
## Uses
#### Direct Use
The model is a classifier that can be used to detect text generated by GPT-2 models.
#### Downstream Use
The model's developers have stated that they developed and released the model to help with research related to synthetic text generation, so the model could potentially be used for downstream tasks related to synthetic text generation. See the [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf) for further discussion.
#### Misuse and Out-of-scope Use
The model should not be used to intentionally create hostile or alienating environments for people. In addition, the model developers discuss the risk of adversaries using the model to better evade detection in their [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf), suggesting that using the model for evading detection or for supporting efforts to evade detection would be a misuse of the model.
## Risks, Limitations and Biases
**CONTENT WARNING: Readers should be aware this section may contain content that is disturbing, offensive, and can propagate historical and current stereotypes.**
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
#### Risks and Limitations
In their [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf), the model developers discuss the risk that the model may be used by bad actors to develop capabilities for evading detection, though one purpose of releasing the model is to help improve detection research.
In a related [blog post](https://openai.com/blog/gpt-2-1-5b-release/), the model developers also discuss the limitations of automated methods for detecting synthetic text and the need to pair automated detection tools with other, non-automated approaches. They write:
> We conducted in-house detection research and developed a detection model that has detection rates of ~95% for detecting 1.5B GPT-2-generated text. We believe this is not high enough accuracy for standalone detection and needs to be paired with metadata-based approaches, human judgment, and public education to be more effective.
The model developers also [report](https://openai.com/blog/gpt-2-1-5b-release/) finding that classifying content from larger models is more difficult, suggesting that detection with automated tools like this model will be increasingly difficult as model sizes increase. The authors find that training detector models on the outputs of larger models can improve accuracy and robustness.
#### Bias
Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)). Predictions generated by RoBERTa large and GPT-2 1.5B (which this model is built/fine-tuned on) can include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups (see the [RoBERTa large](https://huggingface.co/roberta-large) and [GPT-2 XL](https://huggingface.co/gpt2-xl) model cards for more information). The developers of this model discuss these issues further in their [paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf).
## Training
#### Training Data
The model is a sequence classifier based on RoBERTa large (see the [RoBERTa large model card](https://huggingface.co/roberta-large) for more details on the RoBERTa large training data) and then fine-tuned using the outputs of the 1.5B GPT-2 model (available [here](https://github.com/openai/gpt-2-output-dataset)).
#### Training Procedure
The model developers write that:
> We based a sequence classifier on RoBERTaLARGE (355 million parameters) and fine-tuned it to classify the outputs from the 1.5B GPT-2 model versus WebText, the dataset we used to train the GPT-2 model.
They later state:
> To develop a robust detector model that can accurately classify generated texts regardless of the sampling method, we performed an analysis of the model’s transfer performance.
See the [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf) for further details on the training procedure.
## Evaluation
The following evaluation information is extracted from the [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf).
#### Testing Data, Factors and Metrics
The model is intended to be used for detecting text generated by GPT-2 models, so the model developers test the model on text datasets, measuring accuracy by:
> testing 510-token test examples comprised of 5,000 samples from the WebText dataset and 5,000 samples generated by a GPT-2 model, which were not used during the training.
#### Results
The model developers [find](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf):
> Our classifier is able to detect 1.5 billion parameter GPT-2-generated text with approximately 95% accuracy...The model’s accuracy depends on sampling methods used when generating outputs, like temperature, Top-K, and nucleus sampling ([Holtzman et al., 2019](https://arxiv.org/abs/1904.09751). Nucleus sampling outputs proved most difficult to correctly classify, but a detector trained using nucleus sampling transfers well across other sampling methods. As seen in Figure 1 [in the paper], we found consistently high accuracy when trained on nucleus sampling.
See the [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf), Figure 1 (on page 14) and Figure 2 (on page 16) for full results.
## Environmental Impact
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:** Unknown
- **Hours used:** Unknown
- **Cloud Provider:** Unknown
- **Compute Region:** Unknown
- **Carbon Emitted:** Unknown
## Technical Specifications
The model developers write that:
See the [associated paper](https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf) for further details on the modeling architecture and training details.
## Citation Information
```bibtex
@article{solaiman2019release,
title={Release strategies and the social impacts of language models},
author={Solaiman, Irene and Brundage, Miles and Clark, Jack and Askell, Amanda and Herbert-Voss, Ariel and Wu, Jeff and Radford, Alec and Krueger, Gretchen and Kim, Jong Wook and Kreps, Sarah and others},
journal={arXiv preprint arXiv:1908.09203},
year={2019}
}
```
APA:
- Solaiman, I., Brundage, M., Clark, J., Askell, A., Herbert-Voss, A., Wu, J., ... & Wang, J. (2019). Release strategies and the social impacts of language models. arXiv preprint arXiv:1908.09203.
## Model Card Authors
This model card was written by the team at Hugging Face.
## How to Get Started with the Model
More information needed |
vinai/vinai-translate-vi2en | 89ca166856afab4610d0fcc4bff495940b5200ad | 2022-07-06T07:19:15.000Z | [
"pytorch",
"tf",
"mbart",
"text2text-generation",
"transformers",
"autotrain_compatible"
] | text2text-generation | false | vinai | null | vinai/vinai-translate-vi2en | 928 | null | transformers | 1,817 | # A Vietnamese-English Neural Machine Translation System
Our pre-trained VinAI Translate models `vinai/vinai-translate-vi2en` and `vinai/vinai-translate-en2vi` are state-of-the-art text translation models for Vietnamese-to-English and English-to-Vietnamese, respectively. The general architecture and experimental results of VinAI Translate can be found in [our paper](https://openreview.net/forum?id=CRg-RaxKnai):
@inproceedings{vinaitranslate,
title = {{A Vietnamese-English Neural Machine Translation System}},
author = {Thien Hai Nguyen and Tuan-Duy H. Nguyen and Duy Phung and Duy Tran-Cong Nguyen and Hieu Minh Tran and Manh Luong and Tin Duy Vo and Hung Hai Bui and Dinh Phung and Dat Quoc Nguyen},
booktitle = {Proceedings of the 23rd Annual Conference of the International Speech Communication Association: Show and Tell (INTERSPEECH)},
year = {2022}
}
Please **CITE** our paper whenever the pre-trained models or the system are used to help produce published results or incorporated into other software.
For further information or requests, please go to [VinAI Translate's homepage](https://github.com/VinAIResearch/VinAI_Translate)! |
thu-coai/LongLM-small | efadc09553dbe3bfc965dc82aeb9be3d1f224212 | 2021-11-24T05:12:01.000Z | [
"pytorch",
"t5",
"text2text-generation",
"zh",
"arxiv:2108.12960",
"transformers",
"lm-head",
"autotrain_compatible"
] | text2text-generation | false | thu-coai | null | thu-coai/LongLM-small | 925 | 1 | transformers | 1,818 | ---
language:
- zh
thumbnail: http://coai.cs.tsinghua.edu.cn/coai/img/logo.png?v=13923
tags:
- pytorch
- lm-head
- zh
datasets:
metrics:
widget:
- text: "小咕噜对靳司寒完全是个自来熟，小家伙爬进他怀里小手搂着他的脖子，奶声奶气的要求：“靳蜀黎,你给咕噜讲故事好不好？”讲故事？童话故事吗？“我不会。”小家伙明显不信。嘟着小嘴大眼汪汪的盯着他，“哼。”小家伙轻轻哼了一声,靳司寒默了半晌，<extra_id_1>"
- text: "美女亲自打招呼，这可是破天荒第一次，之前不管他献多少次殷勤，美女<extra_id_1>甩他，难道今天真是老天<extra_id_2>不敢<extra_id_3>的兄连滚带爬的来到<extra_id_4>身边队友都带着艳<extra_id_5>他，<extra_id_6>连计算机系的那票球友都在那儿不住地偷看MAGGIE，这种感觉真<extra_id_7>毙了！"
inference:
parameters:
top_p: 0.9
---
## LongLM
### 1. Parameters
| Versions | $d_m$ | $d_{ff}$ | $d_{kv}$ | $n_h$ | $n_e/n_d$ | \# P |
| ------------ | ----- | -------- | -------- | ----- | --------- | ---- |
| LongLM-small | 512 | 2,048 | 64 | 8 | 6/6 | 60M |
| LongLM-base | 768 | 3,072 | 64 | 12 | 12/12 | 223M |
| LongLM-large | 1,536 | 3,072 | 64 | 12 | 24/32 | 1B |
- $d_m$: the dimension of hidden states
- $d_{ff}$: the dimension of feed forward layers
- $d_{kv}$: the dimension of the keys/values in the self-attention layers
- $n_h$: the number of attention heads
- $n_e$: the number of hidden layers of the encoder
- $n_d$: the number of hidden layers of the decoder
- \#P: the number of parameters
### 2. Pretraining Tasks
Encoder-decoder models are trained typically by maximizing the likelihood of the target output given an input. To improve the capacities of both the encoder and decoder, we propose to train LongLM with two pretraining tasks including text infilling (Raffel et al., 2020) and conditional continuation (Radford et al., 2019). For the first task, the input is a text where a number of spans are sampled and replaced by special tokens with unique IDs, while the output is the spans delimited by the special tokens used in the input. The lengths of masked spans are drawn from a Poisson distribution with λ=3 and all masked tokens compress 15% of the original texts. As for the second task, the input and output are respectively the front and back half of a text, which is split into two parts randomly.
### 3. Pretraining Data
We collect 120G novels as the pretraining data for LongLM.
### 4. Checkpoints
1. **Model Loading:**
```python\
from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained('LongLM-large')
model = T5ForConditionalGeneration.from_pretrained('LongLM-large')
```
2. **Generation:**
```python
input_ids = tokenizer("小咕噜对，<extra_id_1>",return_tensors="pt", padding=True, truncation=True, max_length=512).input_ids.to(device)
gen = model.generate(input_ids, do_sample=True, decoder_start_token_id=1, top_p=0.9, max_length=512)
```
### 5. Dependencies
```
datasets 1.6.2
deepspeed 0.3.16
huggingface-hub 0.0.8
jieba 0.42.1
jsonlines 2.0.0
nltk 3.5
numpy 1.19.5
pytorch-lightning 1.2.0
regex 2020.11.13
rouge 1.0.1
rouge-score 0.0.4
sacrebleu 1.5.0
scipy 1.5.4
sentencepiece 0.1.95
tokenizers 0.10.1
torch 1.8.1
torchaudio 0.8.0
torchmetrics 0.2.0
torchvision 0.9.0
transformers 4.6.1
```
### 6. Contributers
[Jian Guan](https://jianguanthu.github.io/) at [thu-coai](http://coai.cs.tsinghua.edu.cn/)
## Citation
```txt
@misc{guan2021lot,
title={LOT: A Benchmark for Evaluating Chinese Long Text Understanding and Generation},
author={Jian Guan and Zhuoer Feng and Yamei Chen and Ruilin He and Xiaoxi Mao and Changjie Fan and Minlie Huang},
year={2021},
eprint={2108.12960},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
|
IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese | 7022713ba0b7754e6017d9acfa18d894fdaad847 | 2022-07-30T02:15:06.000Z | [
"pytorch",
"pegasus",
"text2text-generation",
"zh",
"transformers",
"summarization",
"autotrain_compatible"
] | summarization | false | IDEA-CCNL | null | IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese | 924 | 2 | transformers | 1,819 | ---
language: zh
tags:
- summarization
inference: False
---
IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese model (Chinese) has 523M million parameter, pretrained on 180G Chinese data with GSG task which is stochastically sample important sentences with sampled gap sentence ratios by 25%. The pretraining task just as same as the paper PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization mentioned.
Different from the English version of pegasus, considering that the Chinese sentence piece is unstable, we use jieba and Bertokenizer as the tokenizer in chinese pegasus model.
After pre-training, We use 8 summary datasets which we collect on the internet to do the supervised training. The 8 datasets include education_data, new2016zh_data, nlpcc, shence_data, sohu_data, thucnews_data and weibo_data, four million training samples in all.
And then we use lcsts datasets to train and evalute this model, the results are shown below.
| datasets | rouge-1 | rouge-2 | rouge-L |
| ---- | ---- | ---- | ---- |
| LCSTS | 48.00 | 35.24 | 44.70 |
We also trained a base model, available with [IDEA-CCNL/Randeng-Pegasus-238M-Summary-Chinese](https://huggingface.co/IDEA-CCNL/Randeng-Pegasus-238M-Summary-Chinese)
Task: Summarization
## Usage
```python
from transformers import PegasusForConditionalGeneration
# Need to download tokenizers_pegasus.py and other Python script from Fengshenbang-LM github repo in advance,
# or you can download tokenizers_pegasus.py and data_utils.py in https://huggingface.co/IDEA-CCNL/Randeng_Pegasus_523M/tree/main
# Strongly recommend you git clone the Fengshenbang-LM repo:
# 1. git clone https://github.com/IDEA-CCNL/Fengshenbang-LM
# 2. cd Fengshenbang-LM/fengshen/examples/pegasus/
# and then you will see the tokenizers_pegasus.py and data_utils.py which are needed by pegasus model
from tokenizers_pegasus import PegasusTokenizer
model = PegasusForConditionalGeneration.from_pretrained("IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese")
tokenizer = PegasusTokenizer.from_pretrained("IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese")
text = "据微信公众号“界面”报道，4日上午10点左右，中国发改委反垄断调查小组突击查访奔驰上海办事处，调取数据材料，并对多名奔驰高管进行了约谈。截止昨日晚9点，包括北京梅赛德斯-奔驰销售服务有限公司东区总经理在内的多名管理人员仍留在上海办公室内"
inputs = tokenizer(text, max_length=1024, return_tensors="pt")
# Generate Summary
summary_ids = model.generate(inputs["input_ids"])
tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
# model Output: 反垄断调查小组突击查访奔驰上海办事处，对多名奔驰高管进行约谈
```
## Citation
If you find the resource is useful, please cite the following website in your paper.
```
@misc{Fengshenbang-LM,
title={Fengshenbang-LM},
author={IDEA-CCNL},
year={2022},
howpublished={\url{https://github.com/IDEA-CCNL/Fengshenbang-LM}},
}
``` |
BSC-TeMU/roberta-base-biomedical-es | 6e457abe0082958dc8cb7762c9ec8ed8b8a7b2c0 | 2021-10-21T10:28:29.000Z | [
"pytorch",
"roberta",
"fill-mask",
"es",
"arxiv:2109.03570",
"arxiv:2109.07765",
"transformers",
"biomedical",
"spanish",
"license:apache-2.0",
"autotrain_compatible"
] | fill-mask | false | BSC-TeMU | null | BSC-TeMU/roberta-base-biomedical-es | 923 | 3 | transformers | 1,820 | ---
language:
- es
tags:
- biomedical
- spanish
license: apache-2.0
metrics:
- ppl
widget:
- text: "El único antecedente personal a reseñar era la <mask> arterial."
- text: "Las radiologías óseas de cuerpo entero no detectan alteraciones <mask>, ni alteraciones vertebrales."
- text: "En el <mask> toraco-abdómino-pélvico no se encontraron hallazgos patológicos de interés."
---
**⚠️NOTICE⚠️: THIS MODEL HAS BEEN MOVED TO THE FOLLOWING URL AND WILL SOON BE REMOVED:** https://huggingface.co/PlanTL-GOB-ES/roberta-base-biomedical-es
# Biomedical language model for Spanish
Biomedical pretrained language model for Spanish. For more details about the corpus, the pretraining and the evaluation, check the official [repository](https://github.com/PlanTL-SANIDAD/lm-biomedical-clinical-es) and read our [preprint](https://arxiv.org/abs/2109.03570) "_Carrino, C. P., Armengol-Estapé, J., Gutiérrez-Fandiño, A., Llop-Palao, J., Pàmies, M., Gonzalez-Agirre, A., & Villegas, M. (2021). Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario._".
## Tokenization and model pretraining
This model is a [RoBERTa-based](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model trained on a
**biomedical** corpus in Spanish collected from several sources (see next section).
The training corpus has been tokenized using a byte version of [Byte-Pair Encoding (BPE)](https://github.com/openai/gpt-2)
used in the original [RoBERTA](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model with a vocabulary size of 52,000 tokens. The pretraining consists of a masked language model training at the subword level following the approach employed for the RoBERTa base model with the same hyperparameters as in the original work. The training lasted a total of 48 hours with 16 NVIDIA V100 GPUs of 16GB DDRAM, using Adam optimizer with a peak learning rate of 0.0005 and an effective batch size of 2,048 sentences.
## Training corpora and preprocessing
The training corpus is composed of several biomedical corpora in Spanish, collected from publicly available corpora and crawlers.
To obtain a high-quality training corpus, a cleaning pipeline with the following operations has been applied:
- data parsing in different formats
- sentence splitting
- language detection
- filtering of ill-formed sentences
- deduplication of repetitive contents
- keep the original document boundaries
Finally, the corpora are concatenated and further global deduplication among the corpora have been applied.
The result is a medium-size biomedical corpus for Spanish composed of about 963M tokens. The table below shows some basic statistics of the individual cleaned corpora:
| Name | No. tokens | Description |
|-----------------------------------------------------------------------------------------|-------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| [Medical crawler](https://zenodo.org/record/4561970) | 745,705,946 | Crawler of more than 3,000 URLs belonging to Spanish biomedical and health domains. |
| Clinical cases misc. | 102,855,267 | A miscellany of medical content, essentially clinical cases. Note that a clinical case report is a scientific publication where medical practitioners share patient cases and it is different from a clinical note or document. |
| [Scielo](https://github.com/PlanTL-SANIDAD/SciELO-Spain-Crawler) | 60,007,289 | Publications written in Spanish crawled from the Spanish SciELO server in 2017. |
| [BARR2_background](https://temu.bsc.es/BARR2/downloads/background_set.raw_text.tar.bz2) | 24,516,442 | Biomedical Abbreviation Recognition and Resolution (BARR2) containing Spanish clinical case study sections from a variety of clinical disciplines. |
| Wikipedia_life_sciences | 13,890,501 | Wikipedia articles crawled 04/01/2021 with the [Wikipedia API python library](https://pypi.org/project/Wikipedia-API/) starting from the "Ciencias\_de\_la\_vida" category up to a maximum of 5 subcategories. Multiple links to the same articles are then discarded to avoid repeating content. |
| Patents | 13,463,387 | Google Patent in Medical Domain for Spain (Spanish). The accepted codes (Medical Domain) for Json files of patents are: "A61B", "A61C","A61F", "A61H", "A61K", "A61L","A61M", "A61B", "A61P". |
| [EMEA](http://opus.nlpl.eu/download.php?f=EMEA/v3/moses/en-es.txt.zip) | 5,377,448 | Spanish-side documents extracted from parallel corpora made out of PDF documents from the European Medicines Agency. |
| [mespen_Medline](https://zenodo.org/record/3562536#.YTt1fH2xXbR) | 4,166,077 | Spanish-side articles extracted from a collection of Spanish-English parallel corpus consisting of biomedical scientific literature. The collection of parallel resources are aggregated from the MedlinePlus source. |
| PubMed | 1,858,966 | Open-access articles from the PubMed repository crawled in 2017. |
## Evaluation and results
The model has been evaluated on the Named Entity Recognition (NER) using the following datasets:
- [PharmaCoNER](https://zenodo.org/record/4270158): is a track on chemical and drug mention recognition from Spanish medical texts (for more info see: https://temu.bsc.es/pharmaconer/).
- [CANTEMIST](https://zenodo.org/record/3978041#.YTt5qH2xXbQ): is a shared task specifically focusing on named entity recognition of tumor morphology, in Spanish (for more info see: https://zenodo.org/record/3978041#.YTt5qH2xXbQ).
- ICTUSnet: consists of 1,006 hospital discharge reports of patients admitted for stroke from 18 different Spanish hospitals. It contains more than 79,000 annotations for 51 different kinds of variables.
The evaluation results are compared against the [mBERT](https://huggingface.co/bert-base-multilingual-cased) and [BETO](https://huggingface.co/dccuchile/bert-base-spanish-wwm-cased) models:
| F1 - Precision - Recall | roberta-base-biomedical-es | mBERT | BETO |
|---------------------------|----------------------------|-------------------------------|-------------------------|
| PharmaCoNER | **89.48** - **87.85** - **91.18** | 87.46 - 86.50 - 88.46 | 88.18 - 87.12 - 89.28 |
| CANTEMIST | **83.87** - **81.70** - **86.17** | 82.61 - 81.12 - 84.15 | 82.42 - 80.91 - 84.00 |
| ICTUSnet | **88.12** - **85.56** - **90.83** | 86.75 - 83.53 - 90.23 | 85.95 - 83.10 - 89.02 |
## Intended uses & limitations
The model is ready-to-use only for masked language modelling to perform the Fill Mask task (try the inference API or read the next section)
However, the is intended to be fine-tuned on downstream tasks such as Named Entity Recognition or Text Classification.
## Cite
If you use our models, please cite our latest preprint:
```bibtex
@misc{carrino2021biomedical,
title={Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario},
author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Asier Gutiérrez-Fandiño and Joan Llop-Palao and Marc Pàmies and Aitor Gonzalez-Agirre and Marta Villegas},
year={2021},
eprint={2109.03570},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
If you use our Medical Crawler corpus, please cite the preprint:
```bibtex
@misc{carrino2021spanish,
title={Spanish Biomedical Crawled Corpus: A Large, Diverse Dataset for Spanish Biomedical Language Models},
author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Ona de Gibert Bonet and Asier Gutiérrez-Fandiño and Aitor Gonzalez-Agirre and Martin Krallinger and Marta Villegas},
year={2021},
eprint={2109.07765},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
---
## How to use
```python
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")
model = AutoModelForMaskedLM.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")
from transformers import pipeline
unmasker = pipeline('fill-mask', model="BSC-TeMU/roberta-base-biomedical-es")
unmasker("El único antecedente personal a reseñar era la <mask> arterial.")
```
```
# Output
[
{
"sequence": " El único antecedente personal a reseñar era la hipertensión arterial.",
"score": 0.9855039715766907,
"token": 3529,
"token_str": " hipertensión"
},
{
"sequence": " El único antecedente personal a reseñar era la diabetes arterial.",
"score": 0.0039140828885138035,
"token": 1945,
"token_str": " diabetes"
},
{
"sequence": " El único antecedente personal a reseñar era la hipotensión arterial.",
"score": 0.002484665485098958,
"token": 11483,
"token_str": " hipotensión"
},
{
"sequence": " El único antecedente personal a reseñar era la Hipertensión arterial.",
"score": 0.0023484621196985245,
"token": 12238,
"token_str": " Hipertensión"
},
{
"sequence": " El único antecedente personal a reseñar era la presión arterial.",
"score": 0.0008009297889657319,
"token": 2267,
"token_str": " presión"
}
]
``` |
imxly/sentence_rtb3 | ac6aefbe53994837a744006c493e9980874179c7 | 2021-05-19T20:21:34.000Z | [
"pytorch",
"jax",
"bert",
"feature-extraction",
"transformers"
] | feature-extraction | false | imxly | null | imxly/sentence_rtb3 | 923 | null | transformers | 1,821 | Entry not found |
UBC-NLP/AraT5-base | ed49be981b4df4040e83de16fd559e191b87429f | 2022-05-26T18:25:19.000Z | [
"pytorch",
"tf",
"t5",
"ar",
"transformers",
"Arabic T5",
"MSA",
"Twitter",
"Arabic Dialect",
"Arabic Machine Translation",
"Arabic Text Summarization",
"Arabic News Title and Question Generation",
"Arabic Paraphrasing and Transliteration",
"Arabic Code-Switched Translation"
] | null | false | UBC-NLP | null | UBC-NLP/AraT5-base | 922 | 4 | transformers | 1,822 | ---
language:
- ar
tags:
- Arabic T5
- MSA
- Twitter
- Arabic Dialect
- Arabic Machine Translation
- Arabic Text Summarization
- Arabic News Title and Question Generation
- Arabic Paraphrasing and Transliteration
- Arabic Code-Switched Translation
---
# AraT5-base
# AraT5: Text-to-Text Transformers for Arabic Language Generation
<img src="https://huggingface.co/UBC-NLP/AraT5-base/resolve/main/AraT5_CR_new.png" alt="AraT5" width="45%" height="35%" align="right"/>
This is the repository accompanying our paper [AraT5: Text-to-Text Transformers for Arabic Language Understanding and Generation](https://aclanthology.org/2022.acl-long.47/). In this is the repository we Introduce **AraT5<sub>MSA</sub>**, **AraT5<sub>Tweet</sub>**, and **AraT5**: three powerful Arabic-specific text-to-text Transformer based models;
---
# How to use AraT5 models
Below is an example for fine-tuning **AraT5-base** for News Title Generation on the Aranews dataset
``` bash
!python run_trainier_seq2seq_huggingface.py \
--learning_rate 5e-5 \
--max_target_length 128 --max_source_length 128 \
--per_device_train_batch_size 8 --per_device_eval_batch_size 8 \
--model_name_or_path "UBC-NLP/AraT5-base" \
--output_dir "/content/AraT5_FT_title_generation" --overwrite_output_dir \
--num_train_epochs 3 \
--train_file "/content/ARGEn_title_genration_sample_train.tsv" \
--validation_file "/content/ARGEn_title_genration_sample_valid.tsv" \
--task "title_generation" --text_column "document" --summary_column "title" \
--load_best_model_at_end --metric_for_best_model "eval_bleu" --greater_is_better True --evaluation_strategy epoch --logging_strategy epoch --predict_with_generate\
--do_train --do_eval
```
For more details about the fine-tuning example, please read this notebook [](https://github.com/UBC-NLP/araT5/blob/main/examples/Fine_tuning_AraT5.ipynb)
In addition, we release the fine-tuned checkpoint of the News Title Generation (NGT) which is described in the paper. The model available at Huggingface ([UBC-NLP/AraT5-base-title-generation](https://huggingface.co/UBC-NLP/AraT5-base-title-generation)).
For more details, please visit our own [GitHub](https://github.com/UBC-NLP/araT5).
# AraT5 Models Checkpoints
AraT5 Pytorch and TensorFlow checkpoints are available on the Huggingface website for direct download and use ```exclusively for research```. ```For commercial use, please contact the authors via email @ (muhammad.mageed[at]ubc[dot]ca).```
| **Model** | **Link** |
|---------|:------------------:|
| **AraT5-base** | [https://huggingface.co/UBC-NLP/AraT5-base](https://huggingface.co/UBC-NLP/AraT5-base) |
| **AraT5-msa-base** | [https://huggingface.co/UBC-NLP/AraT5-msa-base](https://huggingface.co/UBC-NLP/AraT5-msa-base) |
| **AraT5-tweet-base** | [https://huggingface.co/UBC-NLP/AraT5-tweet-base](https://huggingface.co/UBC-NLP/AraT5-tweet-base) |
| **AraT5-msa-small** | [https://huggingface.co/UBC-NLP/AraT5-msa-small](https://huggingface.co/UBC-NLP/AraT5-msa-small) |
| **AraT5-tweet-small**| [https://huggingface.co/UBC-NLP/AraT5-tweet-small](https://huggingface.co/UBC-NLP/AraT5-tweet-small) |
# BibTex
If you use our models (Arat5-base, Arat5-msa-base, Arat5-tweet-base, Arat5-msa-small, or Arat5-tweet-small ) for your scientific publication, or if you find the resources in this repository useful, please cite our paper as follows (to be updated):
```bibtex
@inproceedings{nagoudi2022_arat5,
@inproceedings{nagoudi-etal-2022-arat5,
title = "{A}ra{T}5: Text-to-Text Transformers for {A}rabic Language Generation",
author = "Nagoudi, El Moatez Billah and
Elmadany, AbdelRahim and
Abdul-Mageed, Muhammad",
booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = may,
year = "2022",
address = "Dublin, Ireland",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.acl-long.47",
pages = "628--647",
abstract = "Transfer learning with a unified Transformer framework (T5) that converts all language problems into a text-to-text format was recently proposed as a simple and effective transfer learning approach. Although a multilingual version of the T5 model (mT5) was also introduced, it is not clear how well it can fare on non-English tasks involving diverse data. To investigate this question, we apply mT5 on a language with a wide variety of dialects{--}Arabic. For evaluation, we introduce a novel benchmark for ARabic language GENeration (ARGEN), covering seven important tasks. For model comparison, we pre-train three powerful Arabic T5-style models and evaluate them on ARGEN. Although pre-trained with {\textasciitilde}49 less data, our new models perform significantly better than mT5 on all ARGEN tasks (in 52 out of 59 test sets) and set several new SOTAs. Our models also establish new SOTA on the recently-proposed, large Arabic language understanding evaluation benchmark ARLUE (Abdul-Mageed et al., 2021). Our new models are publicly available. We also link to ARGEN datasets through our repository: https://github.com/UBC-NLP/araT5.",
}
## Acknowledgments
We gratefully acknowledge support from the Natural Sciences and Engineering Research Council of Canada, the Social Sciences and Humanities Research Council of Canada, Canadian Foundation for Innovation, [ComputeCanada](www.computecanada.ca) and [UBC ARC-Sockeye](https://doi.org/10.14288/SOCKEYE). We also thank the [Google TensorFlow Research Cloud (TFRC)](https://www.tensorflow.org/tfrc) program for providing us with free TPU access.
|
jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | 6ec4ebf736ed4a6cb093a2b8665e16d55ba0fcc6 | 2022-07-27T23:38:25.000Z | [
"pytorch",
"jax",
"wav2vec2",
"automatic-speech-recognition",
"pt",
"dataset:common_voice",
"dataset:mozilla-foundation/common_voice_6_0",
"transformers",
"audio",
"hf-asr-leaderboard",
"mozilla-foundation/common_voice_6_0",
"robust-speech-event",
"speech",
"xlsr-fine-tuning-week",
"license:apache-2.0",
"model-index"
] | automatic-speech-recognition | false | jonatasgrosman | null | jonatasgrosman/wav2vec2-large-xlsr-53-portuguese | 922 | 7 | transformers | 1,823 | ---
language: pt
license: apache-2.0
datasets:
- common_voice
- mozilla-foundation/common_voice_6_0
metrics:
- wer
- cer
tags:
- audio
- automatic-speech-recognition
- hf-asr-leaderboard
- mozilla-foundation/common_voice_6_0
- pt
- robust-speech-event
- speech
- xlsr-fine-tuning-week
model-index:
- name: XLSR Wav2Vec2 Portuguese by Jonatas Grosman
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pt
type: common_voice
args: pt
metrics:
- name: Test WER
type: wer
value: 11.31
- name: Test CER
type: cer
value: 3.74
- name: Test WER (+LM)
type: wer
value: 9.01
- name: Test CER (+LM)
type: cer
value: 3.21
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Robust Speech Event - Dev Data
type: speech-recognition-community-v2/dev_data
args: pt
metrics:
- name: Dev WER
type: wer
value: 42.1
- name: Dev CER
type: cer
value: 17.93
- name: Dev WER (+LM)
type: wer
value: 36.92
- name: Dev CER (+LM)
type: cer
value: 16.88
---
# Fine-tuned XLSR-53 large model for speech recognition in Portuguese
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Portuguese using the train and validation splits of [Common Voice 6.1](https://huggingface.co/datasets/common_voice).
When using this model, make sure that your speech input is sampled at 16kHz.
This model has been fine-tuned thanks to the GPU credits generously given by the [OVHcloud](https://www.ovhcloud.com/en/public-cloud/ai-training/) :)
The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint
## Usage
The model can be used directly (without a language model) as follows...
Using the [HuggingSound](https://github.com/jonatasgrosman/huggingsound) library:
```python
from huggingsound import SpeechRecognitionModel
model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-large-xlsr-53-portuguese")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = model.transcribe(audio_paths)
```
Writing your own inference script:
```python
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pt"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-portuguese"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
predicted_sentences = processor.batch_decode(predicted_ids)
for i, predicted_sentence in enumerate(predicted_sentences):
print("-" * 100)
print("Reference:", test_dataset[i]["sentence"])
print("Prediction:", predicted_sentence)
```
| Reference | Prediction |
| ------------- | ------------- |
| NEM O RADAR NEM OS OUTROS INSTRUMENTOS DETECTARAM O BOMBARDEIRO STEALTH. | NEMHUM VADAN OS OLTWES INSTRUMENTOS DE TTÉÃN UM BOMBERDEIRO OSTER |
| PEDIR DINHEIRO EMPRESTADO ÀS PESSOAS DA ALDEIA | E DIR ENGINHEIRO EMPRESTAR AS PESSOAS DA ALDEIA |
| OITO | OITO |
| TRANCÁ-LOS | TRANCAUVOS |
| REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA | REALIZAR UMA INVESTIGAÇÃO PARA RESOLVER O PROBLEMA |
| O YOUTUBE AINDA É A MELHOR PLATAFORMA DE VÍDEOS. | YOUTUBE AINDA É A MELHOR PLATAFOMA DE VÍDEOS |
| MENINA E MENINO BEIJANDO NAS SOMBRAS | MENINA E MENINO BEIJANDO NAS SOMBRAS |
| EU SOU O SENHOR | EU SOU O SENHOR |
| DUAS MULHERES QUE SENTAM-SE PARA BAIXO LENDO JORNAIS. | DUAS MIERES QUE SENTAM-SE PARA BAICLANE JODNÓI |
| EU ORIGINALMENTE ESPERAVA | EU ORIGINALMENTE ESPERAVA |
## Evaluation
1. To evaluate on `mozilla-foundation/common_voice_6_0` with split `test`
```bash
python eval.py --model_id jonatasgrosman/wav2vec2-large-xlsr-53-portuguese --dataset mozilla-foundation/common_voice_6_0 --config pt --split test
```
2. To evaluate on `speech-recognition-community-v2/dev_data`
```bash
python eval.py --model_id jonatasgrosman/wav2vec2-large-xlsr-53-portuguese --dataset speech-recognition-community-v2/dev_data --config pt --split validation --chunk_length_s 5.0 --stride_length_s 1.0
```
## Citation
If you want to cite this model you can use this:
```bibtex
@misc{grosman2021xlsr53-large-portuguese,
title={Fine-tuned {XLSR}-53 large model for speech recognition in {P}ortuguese},
author={Grosman, Jonatas},
howpublished={\url{https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-portuguese}},
year={2021}
}
```
|
tscholak/3vnuv1vf | 18d448ce4c0f85d8cf9c06ffae8e197d10515ec1 | 2022-01-10T21:49:25.000Z | [
"pytorch",
"t5",
"text2text-generation",
"en",
"dataset:spider",
"arxiv:2109.05093",
"transformers",
"text2sql",
"license:apache-2.0",
"autotrain_compatible"
] | text2text-generation | false | tscholak | null | tscholak/3vnuv1vf | 921 | 2 | transformers | 1,824 | ---
language:
- en
thumbnail: "https://repository-images.githubusercontent.com/401779782/c2f46be5-b74b-4620-ad64-57487be3b1ab"
tags:
- text2sql
widget:
- "How many singers do we have? | concert_singer | stadium : stadium_id, location, name, capacity, highest, lowest, average | singer : singer_id, name, country, song_name, song_release_year, age, is_male | concert : concert_id, concert_name, theme, stadium_id, year | singer_in_concert : concert_id, singer_id"
license: "apache-2.0"
datasets:
- spider
metrics:
- spider
---
## tscholak/3vnuv1vf
Fine-tuned weights for [PICARD - Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models](https://arxiv.org/abs/2109.05093) based on [t5.1.1.lm100k.large](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#lm-adapted-t511lm100k).
### Training Data
The model has been fine-tuned on the 7000 training examples in the [Spider text-to-SQL dataset](https://yale-lily.github.io/spider). The model solves Spider's zero-shot text-to-SQL translation task, and that means that it can generalize to unseen SQL databases.
### Training Objective
This model was initialized with [t5.1.1.lm100k.large](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#lm-adapted-t511lm100k) and fine-tuned with the text-to-text generation objective.
Questions are always grounded in a database schema, and the model is trained to predict the SQL query that would be used to answer the question. The input to the model is composed of the user's natural language question, the database identifier, and a list of tables and their columns:
```
[question] | [db_id] | [table] : [column] ( [content] , [content] ) , [column] ( ... ) , [...] | [table] : ... | ...
```
The model outputs the database identifier and the SQL query that will be executed on the database to answer the user's question:
```
[db_id] | [sql]
```
### Performance
Out of the box, this model achieves 71.2 % exact-set match accuracy and 74.4 % execution accuracy on the Spider development set.
Using the PICARD constrained decoding method (see [the official PICARD implementation](https://github.com/ElementAI/picard)), the model's performance can be improved to **74.8 %** exact-set match accuracy and **79.2 %** execution accuracy on the Spider development set.
### Usage
Please see [the official repository](https://github.com/ElementAI/picard) for scripts and docker images that support evaluation and serving of this model.
### References
1. [PICARD - Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models](https://arxiv.org/abs/2109.05093)
2. [Official PICARD code](https://github.com/ElementAI/picard)
### Citation
```bibtex
@inproceedings{Scholak2021:PICARD,
author = {Torsten Scholak and Nathan Schucher and Dzmitry Bahdanau},
title = "{PICARD}: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models",
booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2021",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.emnlp-main.779",
pages = "9895--9901",
}
``` |
StanfordAIMI/stanford-deidentifier-base | 41f1cf1c95cb9c25643f625b5aeae663d7e07663 | 2022-07-18T03:38:21.000Z | [
"pytorch",
"bert",
"en",
"dataset:radreports",
"transformers",
"token-classification",
"sequence-tagger-model",
"pubmedbert",
"uncased",
"radiology",
"biomedical",
"license:mit"
] | token-classification | false | StanfordAIMI | null | StanfordAIMI/stanford-deidentifier-base | 921 | 1 | transformers | 1,825 | ---
widget:
- text: "PROCEDURE: Chest xray. COMPARISON: last seen on 1/1/2020 and also record dated of March 1st, 2019. FINDINGS: patchy airspace opacities. IMPRESSION: The results of the chest xray of January 1 2020 are the most concerning ones. The patient was transmitted to another service of UH Medical Center under the responsability of Dr. Perez. We used the system MedClinical data transmitter and sent the data on 2/1/2020, under the ID 5874233. We received the confirmation of Dr Perez. He is reachable at 567-493-1234."
- text: "Dr. Curt Langlotz chose to schedule a meeting on 06/23."
tags:
- token-classification
- sequence-tagger-model
- pytorch
- transformers
- pubmedbert
- uncased
- radiology
- biomedical
datasets:
- radreports
language:
- en
license: mit
---
Stanford de-identifier was trained on a variety of radiology and biomedical documents with the goal of automatising the de-identification process while reaching satisfactory accuracy for use in production. Manuscript in-proceedings. |
cross-encoder/stsb-TinyBERT-L-4 | a0fde64e9dea230cae6957f45eaa3a4685620b01 | 2021-08-05T08:41:47.000Z | [
"pytorch",
"jax",
"bert",
"text-classification",
"transformers",
"license:apache-2.0"
] | text-classification | false | cross-encoder | null | cross-encoder/stsb-TinyBERT-L-4 | 920 | null | transformers | 1,826 | ---
license: apache-2.0
---
# Cross-Encoder for Quora Duplicate Questions Detection
This model was trained using [SentenceTransformers](https://sbert.net) [Cross-Encoder](https://www.sbert.net/examples/applications/cross-encoder/README.html) class.
## Training Data
This model was trained on the [STS benchmark dataset](http://ixa2.si.ehu.eus/stswiki/index.php/STSbenchmark). The model will predict a score between 0 and 1 how for the semantic similarity of two sentences.
## Usage and Performance
Pre-trained models can be used like this:
```
from sentence_transformers import CrossEncoder
model = CrossEncoder('model_name')
scores = model.predict([('Sentence 1', 'Sentence 2'), ('Sentence 3', 'Sentence 4')])
```
The model will predict scores for the pairs `('Sentence 1', 'Sentence 2')` and `('Sentence 3', 'Sentence 4')`.
You can use this model also without sentence_transformers and by just using Transformers ``AutoModel`` class |
Helsinki-NLP/opus-mt-bn-en | 1b349f7c24ee5f832ca19138d23cf78de5869e80 | 2021-01-18T07:51:55.000Z | [
"pytorch",
"marian",
"text2text-generation",
"bn",
"en",
"transformers",
"translation",
"license:apache-2.0",
"autotrain_compatible"
] | translation | false | Helsinki-NLP | null | Helsinki-NLP/opus-mt-bn-en | 919 | null | transformers | 1,827 | ---
language:
- bn
- en
tags:
- translation
license: apache-2.0
---
### ben-eng
* source group: Bengali
* target group: English
* OPUS readme: [ben-eng](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/ben-eng/README.md)
* model: transformer-align
* source language(s): ben
* target language(s): eng
* model: transformer-align
* pre-processing: normalization + SentencePiece (spm32k,spm32k)
* download original weights: [opus-2020-06-17.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/ben-eng/opus-2020-06-17.zip)
* test set translations: [opus-2020-06-17.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/ben-eng/opus-2020-06-17.test.txt)
* test set scores: [opus-2020-06-17.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/ben-eng/opus-2020-06-17.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba-test.ben.eng | 49.7 | 0.641 |
### System Info:
- hf_name: ben-eng
- source_languages: ben
- target_languages: eng
- opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/ben-eng/README.md
- original_repo: Tatoeba-Challenge
- tags: ['translation']
- languages: ['bn', 'en']
- src_constituents: {'ben'}
- tgt_constituents: {'eng'}
- src_multilingual: False
- tgt_multilingual: False
- prepro: normalization + SentencePiece (spm32k,spm32k)
- url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/ben-eng/opus-2020-06-17.zip
- url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/ben-eng/opus-2020-06-17.test.txt
- src_alpha3: ben
- tgt_alpha3: eng
- short_pair: bn-en
- chrF2_score: 0.6409999999999999
- bleu: 49.7
- brevity_penalty: 0.976
- ref_len: 13978.0
- src_name: Bengali
- tgt_name: English
- train_date: 2020-06-17
- src_alpha2: bn
- tgt_alpha2: en
- prefer_old: False
- long_pair: ben-eng
- helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
- transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
- port_machine: brutasse
- port_time: 2020-08-21-14:41 |
google/t5-efficient-base | 6b14ca76d201b73fe751ea16df730c5c999ef736 | 2022-02-15T10:49:53.000Z | [
"pytorch",
"tf",
"jax",
"t5",
"text2text-generation",
"en",
"dataset:c4",
"arxiv:2109.10686",
"transformers",
"deep-narrow",
"license:apache-2.0",
"autotrain_compatible"
] | text2text-generation | false | google | null | google/t5-efficient-base | 917 | 2 | transformers | 1,828 | ---
language:
- en
datasets:
- c4
tags:
- deep-narrow
inference: false
license: apache-2.0
---
# T5-Efficient-BASE (Deep-Narrow version)
T5-Efficient-BASE is a variation of [Google's original T5](https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html) following the [T5 model architecture](https://huggingface.co/docs/transformers/model_doc/t5).
It is a *pretrained-only* checkpoint and was released with the
paper **[Scale Efficiently: Insights from Pre-training and Fine-tuning Transformers](https://arxiv.org/abs/2109.10686)**
by *Yi Tay, Mostafa Dehghani, Jinfeng Rao, William Fedus, Samira Abnar, Hyung Won Chung, Sharan Narang, Dani Yogatama, Ashish Vaswani, Donald Metzler*.
In a nutshell, the paper indicates that a **Deep-Narrow** model architecture is favorable for **downstream** performance compared to other model architectures
of similar parameter count.
To quote the paper:
> We generally recommend a DeepNarrow strategy where the model’s depth is preferentially increased
> before considering any other forms of uniform scaling across other dimensions. This is largely due to
> how much depth influences the Pareto-frontier as shown in earlier sections of the paper. Specifically, a
> tall small (deep and narrow) model is generally more efficient compared to the base model. Likewise,
> a tall base model might also generally more efficient compared to a large model. We generally find
> that, regardless of size, even if absolute performance might increase as we continue to stack layers,
> the relative gain of Pareto-efficiency diminishes as we increase the layers, converging at 32 to 36
> layers. Finally, we note that our notion of efficiency here relates to any one compute dimension, i.e.,
> params, FLOPs or throughput (speed). We report all three key efficiency metrics (number of params,
> FLOPS and speed) and leave this decision to the practitioner to decide which compute dimension to
> consider.
To be more precise, *model depth* is defined as the number of transformer blocks that are stacked sequentially.
A sequence of word embeddings is therefore processed sequentially by each transformer block.
## Details model architecture
This model checkpoint - **t5-efficient-base** - is of model type **Base** with no variations.
It has **222.93** million parameters and thus requires *ca.* **891.73 MB** of memory in full precision (*fp32*)
or **445.86 MB** of memory in half precision (*fp16* or *bf16*).
A summary of the *original* T5 model architectures can be seen here:
| Model | nl (el/dl) | ff | dm | kv | nh | #Params|
| ----| ---- | ---- | ---- | ---- | ---- | ----|
| Tiny | 4/4 | 1024 | 256 | 32 | 4 | 16M|
| Mini | 4/4 | 1536 | 384 | 32 | 8 | 31M|
| Small | 6/6 | 2048 | 512 | 32 | 8 | 60M|
| Base | 12/12 | 3072 | 768 | 64 | 12 | 220M|
| Large | 24/24 | 4096 | 1024 | 64 | 16 | 738M|
| Xl | 24/24 | 16384 | 1024 | 128 | 32 | 3B|
| XXl | 24/24 | 65536 | 1024 | 128 | 128 | 11B|
whereas the following abbreviations are used:
| Abbreviation | Definition |
| ----| ---- |
| nl | Number of transformer blocks (depth) |
| dm | Dimension of embedding vector (output vector of transformers block) |
| kv | Dimension of key/value projection matrix |
| nh | Number of attention heads |
| ff | Dimension of intermediate vector within transformer block (size of feed-forward projection matrix) |
| el | Number of transformer blocks in the encoder (encoder depth) |
| dl | Number of transformer blocks in the decoder (decoder depth) |
| sh | Signifies that attention heads are shared |
| skv | Signifies that key-values projection matrices are tied |
If a model checkpoint has no specific, *el* or *dl* than both the number of encoder- and decoder layers correspond to *nl*.
## Pre-Training
The checkpoint was pretrained on the [Colossal, Cleaned version of Common Crawl (C4)](https://huggingface.co/datasets/c4) for 524288 steps using
the span-based masked language modeling (MLM) objective.
## Fine-Tuning
**Note**: This model is a **pretrained** checkpoint and has to be fine-tuned for practical usage.
The checkpoint was pretrained in English and is therefore only useful for English NLP tasks.
You can follow on of the following examples on how to fine-tune the model:
*PyTorch*:
- [Summarization](https://github.com/huggingface/transformers/tree/master/examples/pytorch/summarization)
- [Question Answering](https://github.com/huggingface/transformers/blob/master/examples/pytorch/question-answering/run_seq2seq_qa.py)
- [Text Classification](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) - *Note*: You will have to slightly adapt the training example here to make it work with an encoder-decoder model.
*Tensorflow*:
- [Summarization](https://github.com/huggingface/transformers/tree/master/examples/tensorflow/summarization)
- [Text Classification](https://github.com/huggingface/transformers/tree/master/examples/tensorflow/text-classification) - *Note*: You will have to slightly adapt the training example here to make it work with an encoder-decoder model.
*JAX/Flax*:
- [Summarization](https://github.com/huggingface/transformers/tree/master/examples/flax/summarization)
- [Text Classification](https://github.com/huggingface/transformers/tree/master/examples/flax/text-classification) - *Note*: You will have to slightly adapt the training example here to make it work with an encoder-decoder model.
## Downstream Performance
TODO: Add table if available
## Computational Complexity
TODO: Add table if available
## More information
We strongly recommend the reader to go carefully through the original paper **[Scale Efficiently: Insights from Pre-training and Fine-tuning Transformers](https://arxiv.org/abs/2109.10686)** to get a more nuanced understanding of this model checkpoint.
As explained in the following [issue](https://github.com/google-research/google-research/issues/986#issuecomment-1035051145), checkpoints including the *sh* or *skv*
model architecture variations have *not* been ported to Transformers as they are probably of limited practical usage and are lacking a more detailed description. Those checkpoints are kept [here](https://huggingface.co/NewT5SharedHeadsSharedKeyValues) as they might be ported potentially in the future. |
M-CLIP/XLM-Roberta-Large-Vit-L-14 | ad70f9333ca5fc97d85b1491b939cf721cd2bad8 | 2022-06-02T23:25:42.000Z | [
"pytorch",
"tf",
"multilingual"
] | null | false | M-CLIP | null | M-CLIP/XLM-Roberta-Large-Vit-L-14 | 917 | null | null | 1,829 | ---
language: multilingual
---
## Multilingual-clip: XLM-Roberta-Large-Vit-L-14
Multilingual-CLIP extends OpenAI's English text encoders to multiple other languages. This model *only* contains the multilingual text encoder. The corresponding image model `ViT-L-14` can be retrieved via instructions found on OpenAI's [CLIP repository on Github](https://github.com/openai/CLIP). We provide a usage example below.
## Requirements
To use both the multilingual text encoder and corresponding image encoder, we need to install the packages [`multilingual-clip`](https://github.com/FreddeFrallan/Multilingual-CLIP) and [`clip`](https://github.com/openai/CLIP).
```
pip install multilingual-clip
pip install git+https://github.com/openai/CLIP.git
```
## Usage
Extracting embeddings from the text encoder can be done in the following way:
```python
from multilingual_clip import pt_multilingual_clip
import transformers
texts = [
'Three blind horses listening to Mozart.',
'Älgen är skogens konung!',
'Wie leben Eisbären in der Antarktis?',
'Вы знали, что все белые медведи левши?'
]
model_name = 'M-CLIP/XLM-Roberta-Large-Vit-L-14'
# Load Model & Tokenizer
model = pt_multilingual_clip.MultilingualCLIP.from_pretrained(model_name)
tokenizer = transformers.AutoTokenizer.from_pretrained(model_name)
embeddings = model.forward(texts, tokenizer)
print("Text features shape:", embeddings.shape)
```
Extracting embeddings from the corresponding image encoder:
```python
import torch
import clip
import requests
from PIL import Image
device = "cuda" if torch.cuda.is_available() else "cpu"
model, preprocess = clip.load("ViT-L/14", device=device)
url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)
image = preprocess(image).unsqueeze(0).to(device)
with torch.no_grad():
image_features = model.encode_image(image)
print("Image features shape:", image_features.shape)
```
## Evaluation results
None of the M-CLIP models have been extensivly evaluated, but testing them on Txt2Img retrieval on the humanly translated MS-COCO dataset, we see the following **R@10** results:
| Name | En | De | Es | Fr | Zh | It | Pl | Ko | Ru | Tr | Jp |
| ----------------------------------|:-----: |:-----: |:-----: |:-----: | :-----: |:-----: |:-----: |:-----: |:-----: |:-----: |:-----: |
| [OpenAI CLIP Vit-B/32](https://github.com/openai/CLIP)| 90.3 | - | - | - | - | - | - | - | - | - | - |
| [OpenAI CLIP Vit-L/14](https://github.com/openai/CLIP)| 91.8 | - | - | - | - | - | - | - | - | - | - |
| [OpenCLIP ViT-B-16+-](https://github.com/openai/CLIP)| 94.3 | - | - | - | - | - | - | - | - | - | - |
| [LABSE Vit-L/14](https://huggingface.co/M-CLIP/LABSE-Vit-L-14)| 91.6 | 89.6 | 89.5 | 89.9 | 88.9 | 90.1 | 89.8 | 80.8 | 85.5 | 89.8 | 73.9 |
| [XLM-R Large Vit-B/32](https://huggingface.co/M-CLIP/XLM-Roberta-Large-Vit-B-32)| 91.8 | 88.7 | 89.1 | 89.4 | 89.3 | 89.8| 91.4 | 82.1 | 86.1 | 88.8 | 81.0 |
| [XLM-R Vit-L/14](https://huggingface.co/M-CLIP/XLM-Roberta-Large-Vit-L-14)| 92.4 | 90.6 | 91.0 | 90.0 | 89.7 | 91.1 | 91.3 | 85.2 | 85.8 | 90.3 | 81.9 |
| [XLM-R Large Vit-B/16+](https://huggingface.co/M-CLIP/XLM-Roberta-Large-Vit-B-16Plus)| **95.0** | **93.0** | **93.6** | **93.1** | **94.0** | **93.1** | **94.4** | **89.0** | **90.0** | **93.0** | **84.2** |
## Training/Model details
Further details about the model training and data can be found in the [model card](https://github.com/FreddeFrallan/Multilingual-CLIP/blob/main/larger_mclip.md). |
RuRI/Talkmodel01 | bd6f30cb7839b3955919959f5d58ebca366563f8 | 2021-09-17T00:34:28.000Z | [
"pytorch",
"gpt2",
"text-generation",
"transformers"
] | text-generation | false | RuRI | null | RuRI/Talkmodel01 | 911 | null | transformers | 1,830 | Entry not found |
mrm8488/mobilebert-finetuned-pos | 54dad38c9125524220389490ce914b9c85d598da | 2021-03-12T08:08:35.000Z | [
"pytorch",
"rust",
"mobilebert",
"token-classification",
"en",
"transformers",
"pos",
"license:mit",
"autotrain_compatible"
] | token-classification | false | mrm8488 | null | mrm8488/mobilebert-finetuned-pos | 911 | 4 | transformers | 1,831 | ---
language: en
tags:
- mobilebert
- pos
license: mit
---
|
pszemraj/led-large-book-summary | 16ef34e1d8d5c43c1e5025463848f30531a12077 | 2022-07-21T09:03:04.000Z | [
"pytorch",
"led",
"text2text-generation",
"en",
"dataset:kmfoda/booksum",
"arxiv:2105.08209",
"transformers",
"summarization",
"summary",
"longformer",
"booksum",
"long-document",
"long-form",
"license:apache-2.0",
"model-index",
"autotrain_compatible"
] | summarization | false | pszemraj | null | pszemraj/led-large-book-summary | 911 | 3 | transformers | 1,832 | ---
language:
- en
tags:
- summarization
- led
- summary
- longformer
- booksum
- long-document
- long-form
license: apache-2.0
datasets:
- kmfoda/booksum
metrics:
- rouge
widget:
- text: large earthquakes along a given fault segment do not occur at random intervals
because it takes time to accumulate the strain energy for the rupture. The rates
at which tectonic plates move and accumulate strain at their boundaries are approximately
uniform. Therefore, in first approximation, one may expect that large ruptures
of the same fault segment will occur at approximately constant time intervals.
If subsequent main shocks have different amounts of slip across the fault, then
the recurrence time may vary, and the basic idea of periodic mainshocks must be
modified. For great plate boundary ruptures the length and slip often vary by
a factor of 2. Along the southern segment of the San Andreas fault the recurrence
interval is 145 years with variations of several decades. The smaller the standard
deviation of the average recurrence interval, the more specific could be the long
term prediction of a future mainshock.
example_title: earthquakes
- text: " A typical feed-forward neural field algorithm. Spatiotemporal coordinates\
\ are fed into a neural network that predicts values in the reconstructed domain.\
\ Then, this domain is mapped to the sensor domain where sensor measurements are\
\ available as supervision. Class and Section Problems Addressed Generalization\
\ (Section 2) Inverse problems, ill-posed problems, editability; symmetries. Hybrid\
\ Representations (Section 3) Computation & memory efficiency, representation\
\ capacity, editability: Forward Maps (Section 4) Inverse problems Network Architecture\
\ (Section 5) Spectral bias, integration & derivatives. Manipulating Neural Fields\
\ (Section 6) Edit ability, constraints, regularization. Table 2: The five classes\
\ of techniques in the neural field toolbox each addresses problems that arise\
\ in learning, inference, and control. (Section 3). We can supervise reconstruction\
\ via differentiable forward maps that transform Or project our domain (e.g, 3D\
\ reconstruction via 2D images; Section 4) With appropriate network architecture\
\ choices, we can overcome neural network spectral biases (blurriness) and efficiently\
\ compute derivatives and integrals (Section 5). Finally, we can manipulate neural\
\ fields to add constraints and regularizations, and to achieve editable representations\
\ (Section 6). Collectively, these classes constitute a 'toolbox' of techniques\
\ to help solve problems with neural fields There are three components in a conditional\
\ neural field: (1) An encoder or inference function \u20AC that outputs the conditioning\
\ latent variable 2 given an observation 0 E(0) =2. 2 is typically a low-dimensional\
\ vector, and is often referred to aS a latent code Or feature code_ (2) A mapping\
\ function 4 between Z and neural field parameters O: Y(z) = O; (3) The neural\
\ field itself $. The encoder \u20AC finds the most probable z given the observations\
\ O: argmaxz P(2/0). The decoder maximizes the inverse conditional probability\
\ to find the most probable 0 given Z: arg- max P(Olz). We discuss different encoding\
\ schemes with different optimality guarantees (Section 2.1.1), both global and\
\ local conditioning (Section 2.1.2), and different mapping functions Y (Section\
\ 2.1.3) 2. Generalization Suppose we wish to estimate a plausible 3D surface\
\ shape given a partial or noisy point cloud. We need a suitable prior over the\
\ sur- face in its reconstruction domain to generalize to the partial observations.\
\ A neural network expresses a prior via the function space of its architecture\
\ and parameters 0, and generalization is influenced by the inductive bias of\
\ this function space (Section 5)."
example_title: scientific paper
- text: ' the big variety of data coming from diverse sources is one of the key properties
of the big data phenomenon. It is, therefore, beneficial to understand how data
is generated in various environments and scenarios, before looking at what should
be done with this data and how to design the best possible architecture to accomplish
this The evolution of IT architectures, described in Chapter 2, means that the
data is no longer processed by a few big monolith systems, but rather by a group
of services In parallel to the processing layer, the underlying data storage has
also changed and became more distributed This, in turn, required a significant
paradigm shift as the traditional approach to transactions (ACID) could no longer
be supported. On top of this, cloud computing is becoming a major approach with
the benefits of reducing costs and providing on-demand scalability but at the
same time introducing concerns about privacy, data ownership, etc In the meantime
the Internet continues its exponential growth: Every day both structured and unstructured
data is published and available for processing: To achieve competitive advantage
companies have to relate their corporate resources to external services, e.g.
financial markets, weather forecasts, social media, etc While several of the sites
provide some sort of API to access the data in a more orderly fashion; countless
sources require advanced web mining and Natural Language Processing (NLP) processing
techniques: Advances in science push researchers to construct new instruments
for observing the universe O conducting experiments to understand even better
the laws of physics and other domains. Every year humans have at their disposal
new telescopes, space probes, particle accelerators, etc These instruments generate
huge streams of data, which need to be stored and analyzed. The constant drive
for efficiency in the industry motivates the introduction of new automation techniques
and process optimization: This could not be done without analyzing the precise
data that describe these processes. As more and more human tasks are automated,
machines provide rich data sets, which can be analyzed in real-time to drive efficiency
to new levels. Finally, it is now evident that the growth of the Internet of Things
is becoming a major source of data. More and more of the devices are equipped
with significant computational power and can generate a continuous data stream
from their sensors. In the subsequent sections of this chapter, we will look at
the domains described above to see what they generate in terms of data sets. We
will compare the volumes but will also look at what is characteristic and important
from their respective points of view. 3.1 The Internet is undoubtedly the largest
database ever created by humans. While several well described; cleaned, and structured
data sets have been made available through this medium, most of the resources
are of an ambiguous, unstructured, incomplete or even erroneous nature. Still,
several examples in the areas such as opinion mining, social media analysis, e-governance,
etc, clearly show the potential lying in these resources. Those who can successfully
mine and interpret the Internet data can gain unique insight and competitive advantage
in their business An important area of data analytics on the edge of corporate
IT and the Internet is Web Analytics.'
example_title: data science textbook
- text: "Transformer-based models have shown to be very useful for many NLP tasks.\
\ However, a major limitation of transformers-based models is its O(n^2)O(n 2)\
\ time & memory complexity (where nn is sequence length). Hence, it's computationally\
\ very expensive to apply transformer-based models on long sequences n > 512n>512.\
\ Several recent papers, e.g. Longformer, Performer, Reformer, Clustered attention\
\ try to remedy this problem by approximating the full attention matrix. You can\
\ checkout \U0001F917's recent blog post in case you are unfamiliar with these\
\ models.\nBigBird (introduced in paper) is one of such recent models to address\
\ this issue. BigBird relies on block sparse attention instead of normal attention\
\ (i.e. BERT's attention) and can handle sequences up to a length of 4096 at a\
\ much lower computational cost compared to BERT. It has achieved SOTA on various\
\ tasks involving very long sequences such as long documents summarization, question-answering\
\ with long contexts.\nBigBird RoBERTa-like model is now available in \U0001F917\
Transformers. The goal of this post is to give the reader an in-depth understanding\
\ of big bird implementation & ease one's life in using BigBird with \U0001F917\
Transformers. But, before going into more depth, it is important to remember that\
\ the BigBird's attention is an approximation of BERT's full attention and therefore\
\ does not strive to be better than BERT's full attention, but rather to be more\
\ efficient. It simply allows to apply transformer-based models to much longer\
\ sequences since BERT's quadratic memory requirement quickly becomes unbearable.\
\ Simply put, if we would have \u221E compute & \u221E time, BERT's attention\
\ would be preferred over block sparse attention (which we are going to discuss\
\ in this post).\nIf you wonder why we need more compute when working with longer\
\ sequences, this blog post is just right for you!\nSome of the main questions\
\ one might have when working with standard BERT-like attention include:\nDo all\
\ tokens really have to attend to all other tokens? Why not compute attention\
\ only over important tokens? How to decide what tokens are important? How to\
\ attend to just a few tokens in a very efficient way? In this blog post, we will\
\ try to answer those questions.\nWhat tokens should be attended to? We will give\
\ a practical example of how attention works by considering the sentence 'BigBird\
\ is now available in HuggingFace for extractive question answering'. In BERT-like\
\ attention, every word would simply attend to all other tokens.\nLet's think\
\ about a sensible choice of key tokens that a queried token actually only should\
\ attend to by writing some pseudo-code. Will will assume that the token available\
\ is queried and build a sensible list of key tokens to attend to.\n>>> # let's\
\ consider following sentence as an example >>> example = ['BigBird', 'is', 'now',\
\ 'available', 'in', 'HuggingFace', 'for', 'extractive', 'question', 'answering']\n\
>>> # further let's assume, we're trying to understand the representation of 'available'\
\ i.e. >>> query_token = 'available' >>> # We will initialize an empty `set` and\
\ fill up the tokens of our interest as we proceed in this section. >>> key_tokens\
\ = [] # => currently 'available' token doesn't have anything to attend Nearby\
\ tokens should be important because, in a sentence (sequence of words), the current\
\ word is highly dependent on neighboring past & future tokens. This intuition\
\ is the idea behind the concept of sliding attention."
example_title: bigbird blog intro
- text: 'The majority of available text summarization datasets include short-form
source documents that lack long-range causal and temporal dependencies, and often
contain strong layout and stylistic biases. While relevant, such datasets will
offer limited challenges for future generations of text summarization systems.
We address these issues by introducing BookSum, a collection of datasets for long-form
narrative summarization. Our dataset covers source documents from the literature
domain, such as novels, plays and stories, and includes highly abstractive, human
written summaries on three levels of granularity of increasing difficulty: paragraph-,
chapter-, and book-level. The domain and structure of our dataset poses a unique
set of challenges for summarization systems, which include: processing very long
documents, non-trivial causal and temporal dependencies, and rich discourse structures.
To facilitate future work, we trained and evaluated multiple extractive and abstractive
summarization models as baselines for our dataset.'
example_title: BookSum Abstract
inference:
parameters:
max_length: 64
min_length: 8
no_repeat_ngram_size: 3
early_stopping: true
repetition_penalty: 3.5
length_penalty: 0.3
encoder_no_repeat_ngram_size: 3
num_beams: 4
model-index:
- name: pszemraj/led-large-book-summary
results:
- task:
type: summarization
name: Summarization
dataset:
name: kmfoda/booksum
type: kmfoda/booksum
config: kmfoda--booksum
split: test
metrics:
- name: ROUGE-1
type: rouge
value: 31.7308
verified: true
- name: ROUGE-2
type: rouge
value: 5.3311
verified: true
- name: ROUGE-L
type: rouge
value: 16.1465
verified: true
- name: ROUGE-LSUM
type: rouge
value: 29.0883
verified: true
- name: loss
type: loss
value: 4.815707206726074
verified: true
- name: gen_len
type: gen_len
value: 154.9036
verified: true
- task:
type: summarization
name: Summarization
dataset:
name: samsum
type: samsum
config: samsum
split: test
metrics:
- name: ROUGE-1
type: rouge
value: 33.4484
verified: true
- name: ROUGE-2
type: rouge
value: 10.4249
verified: true
- name: ROUGE-L
type: rouge
value: 24.5802
verified: true
- name: ROUGE-LSUM
type: rouge
value: 29.8226
verified: true
- name: loss
type: loss
value: 4.176078796386719
verified: true
- name: gen_len
type: gen_len
value: 65.4005
verified: true
---
# Longformer Encoder-Decoder (LED) fine-tuned on Booksum
- This model is a fine-tuned version of [allenai/led-large-16384](https://huggingface.co/allenai/led-large-16384) on the booksum dataset.
- the goal was to create a model that can generalize well and is useful in summarizing lots of text in academic and daily usage.
- [I don't want to read anymore give me a notebook implementation](https://colab.research.google.com/gist/pszemraj/3eba944ddc9fc9a4a1bfb21e83b57620/summarization-token-batching.ipynb)
- all the parameters for generation on the API are the same as [the base model](https://huggingface.co/pszemraj/led-base-book-summary) for easy comparison between versions.
- works well on lots of text, can hand 16384 tokens/batch.
> Note: the API is set to generate a max of 64 tokens for runtime reasons, so the summaries may be truncated (depending on length of input text). For best results use python as below.
---
# Usage - Basics
- it is recommended to use `encoder_no_repeat_ngram_size=3` when calling the pipeline object to improve summary quality.
- this param forces the model to use new vocabulary and create an abstractive summary, otherwise it may compile the best _extractive_ summary from the input provided.
- create the pipeline object:
```
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
from transformers import pipeline
hf_name = 'pszemraj/led-large-book-summary'
_model = AutoModelForSeq2SeqLM.from_pretrained(
hf_name,
low_cpu_mem_usage=True,
)
_tokenizer = AutoTokenizer.from_pretrained(
hf_name
)
summarizer = pipeline(
"summarization",
model=_model,
tokenizer=_tokenizer
)
```
- put words into the pipeline object:
```
wall_of_text = "your words here"
result = summarizer(
wall_of_text,
min_length=16,
max_length=256,
no_repeat_ngram_size=3,
encoder_no_repeat_ngram_size =3,
clean_up_tokenization_spaces=True,
repetition_penalty=3.7,
num_beams=4,
early_stopping=True,
)
```
- **Important:** To generate the best quality summaries, you should use the global attention mask when decoding, as demonstrated in [this community notebook here](https://colab.research.google.com/drive/12INTTR6n64TzS4RrXZxMSXfrOd9Xzamo?usp=sharing), see the definition of `generate_answer(batch)`.
- If you run into compute constraints, try the base version [`pszemraj/led-base-book-summary`](https://huggingface.co/pszemraj/led-base-book-summary)
## Training and evaluation data
- the [booksum](https://arxiv.org/abs/2105.08209) dataset
- During training, the input text was the text of the `chapter`, and the output was `summary_text`
- Eval results can be found [here](https://huggingface.co/datasets/autoevaluate/autoeval-staging-eval-project-kmfoda__booksum-79c1c0d8-10905463) with metrics on the sidebar.
## Training procedure
- Training completed on the BookSum dataset for 13 total epochs
- **The final four epochs combined the training and validation sets as 'train' in an effort to increase generalization.**
### Training hyperparameters
#### Initial Three Epochs
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
#### In-between Epochs
Unfortunately, don't have all records on-hand for middle epochs, the following should be representative:
- learning_rate: 4e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 16
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 6 (in addition to prior model)
#### Final Two Epochs
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 16
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 2 (in addition to prior model)
### Framework versions
- Transformers 4.19.2
- Pytorch 1.11.0+cu113
- Datasets 2.2.2
- Tokenizers 0.12.1 |
EleutherAI/enformer-official-rough | affe5713ae9017460706a44108289b13c5fee16c | 2022-06-12T20:46:42.000Z | [
"pytorch",
"enformer",
"transformers",
"license:cc-by-4.0"
] | null | false | EleutherAI | null | EleutherAI/enformer-official-rough | 910 | 4 | transformers | 1,833 | ---
license: cc-by-4.0
inference: false
---
# Enformer
Enformer model. It was introduced in the paper [Effective gene expression prediction from sequence by integrating long-range interactions.](https://www.nature.com/articles/s41592-021-01252-x) by Avsec et al. and first released in [this repository](https://github.com/deepmind/deepmind-research/tree/master/enformer).
This repo contains the official weights released by Deepmind, ported over to Pytorch.
## Model description
Enformer is a neural network architecture based on the Transformer that led to greatly increased accuracy in predicting gene expression from DNA sequence.
We refer to the [paper](https://www.nature.com/articles/s41592-021-01252-x) published in Nature for details.
### How to use
Refer to the README of [enformer-pytorch](https://github.com/lucidrains/enformer-pytorch) regarding usage.
### Citation info
```
Avsec, Ž., Agarwal, V., Visentin, D. et al. Effective gene expression prediction from sequence by integrating long-range interactions. Nat Methods 18, 1196–1203 (2021). https://doi.org/10.1038/s41592-021-01252-x
```
|
microsoft/DialogRPT-human-vs-machine | 735475522c2e95409e38a6f7ce714ca72a6bb219 | 2021-05-23T09:16:47.000Z | [
"pytorch",
"gpt2",
"text-classification",
"arxiv:2009.06978",
"transformers"
] | text-classification | false | microsoft | null | microsoft/DialogRPT-human-vs-machine | 908 | null | transformers | 1,834 | # Demo
Please try this [➤➤➤ Colab Notebook Demo (click me!)](https://colab.research.google.com/drive/1cAtfkbhqsRsT59y3imjR1APw3MHDMkuV?usp=sharing)
| Context | Response | `human_vs_machine` score |
| :------ | :------- | :------------: |
| I love NLP! | I'm not sure if it's a good idea. | 0.000 |
| I love NLP! | Me too! | 0.605 |
The `human_vs_machine` score predicts how likely the response is from a human rather than a machine.
# DialogRPT-human-vs-machine
### Dialog Ranking Pretrained Transformers
> How likely a dialog response is upvoted 👍 and/or gets replied 💬?
This is what [**DialogRPT**](https://github.com/golsun/DialogRPT) is learned to predict.
It is a set of dialog response ranking models proposed by [Microsoft Research NLP Group](https://www.microsoft.com/en-us/research/group/natural-language-processing/) trained on 100 + millions of human feedback data.
It can be used to improve existing dialog generation model (e.g., [DialoGPT](https://huggingface.co/microsoft/DialoGPT-medium)) by re-ranking the generated response candidates.
Quick Links:
* [EMNLP'20 Paper](https://arxiv.org/abs/2009.06978/)
* [Dataset, training, and evaluation](https://github.com/golsun/DialogRPT)
* [Colab Notebook Demo](https://colab.research.google.com/drive/1cAtfkbhqsRsT59y3imjR1APw3MHDMkuV?usp=sharing)
We considered the following tasks and provided corresponding pretrained models.
|Task | Description | Pretrained model |
| :------------- | :----------- | :-----------: |
| **Human feedback** | **given a context and its two human responses, predict...**|
| `updown` | ... which gets more upvotes? | [model card](https://huggingface.co/microsoft/DialogRPT-updown) |
| `width`| ... which gets more direct replies? | [model card](https://huggingface.co/microsoft/DialogRPT-width) |
| `depth`| ... which gets longer follow-up thread? | [model card](https://huggingface.co/microsoft/DialogRPT-depth) |
| **Human-like** (human vs fake) | **given a context and one human response, distinguish it with...** |
| `human_vs_rand`| ... a random human response | [model card](https://huggingface.co/microsoft/DialogRPT-human-vs-rand) |
| `human_vs_machine`| ... a machine generated response | this model |
### Contact:
Please create an issue on [our repo](https://github.com/golsun/DialogRPT)
### Citation:
```
@inproceedings{gao2020dialogrpt,
title={Dialogue Response RankingTraining with Large-Scale Human Feedback Data},
author={Xiang Gao and Yizhe Zhang and Michel Galley and Chris Brockett and Bill Dolan},
year={2020},
booktitle={EMNLP}
}
```
|
facebook/levit-128S | dd3c14ead498eab264fe0ed2053dc30940393467 | 2022-06-01T13:20:18.000Z | [
"pytorch",
"levit",
"image-classification",
"dataset:imagenet-1k",
"arxiv:2104.01136",
"transformers",
"vision",
"license:apache-2.0"
] | image-classification | false | facebook | null | facebook/levit-128S | 906 | 1 | transformers | 1,835 | ---
license: apache-2.0
tags:
- vision
- image-classification
datasets:
- imagenet-1k
widget:
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/tiger.jpg
example_title: Tiger
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/teapot.jpg
example_title: Teapot
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/palace.jpg
example_title: Palace
---
# LeViT
LeViT-128S model pre-trained on ImageNet-1k at resolution 224x224. It was introduced in the paper [LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
](https://arxiv.org/abs/2104.01136) by Graham et al. and first released in [this repository](https://github.com/facebookresearch/LeViT).
Disclaimer: The team releasing LeViT did not write a model card for this model so this model card has been written by the Hugging Face team.
## Usage
Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 1,000 ImageNet classes:
```python
from transformers import LevitFeatureExtractor, LevitForImageClassificationWithTeacher
from PIL import Image
import requests
url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)
feature_extractor = LevitFeatureExtractor.from_pretrained('facebook/levit-128S')
model = LevitForImageClassificationWithTeacher.from_pretrained('facebook/levit-128S')
inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits
# model predicts one of the 1000 ImageNet classes
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", model.config.id2label[predicted_class_idx])
``` |
PlanTL-GOB-ES/bsc-bio-es | 623c437d1f056466142fd2e73b4e905dc2ef07ff | 2022-04-11T11:02:40.000Z | [
"pytorch",
"roberta",
"fill-mask",
"es",
"transformers",
"biomedical",
"clinical",
"spanish",
"license:apache-2.0",
"autotrain_compatible"
] | fill-mask | false | PlanTL-GOB-ES | null | PlanTL-GOB-ES/bsc-bio-es | 905 | null | transformers | 1,836 | ---
language:
- es
tags:
- biomedical
- clinical
- spanish
license: apache-2.0
metrics:
- ppl
widget:
- text: "El único antecedente personal a reseñar era la <mask> arterial."
- text: "Las radiologías óseas de cuerpo entero no detectan alteraciones <mask>, ni alteraciones vertebrales."
- text: "En el <mask> toraco-abdómino-pélvico no se encontraron hallazgos patológicos de interés."
---
# Biomedical language model for Spanish
Biomedical pretrained language model for Spanish. For more details about the corpus, the pretraining and the evaluation, check the official [repository](https://github.com/PlanTL-GOB-ES/lm-biomedical-clinical-es).
## Tokenization and model pretraining
This model is a [RoBERTa-based](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model trained on a
**biomedical** corpus in Spanish collected from several sources (see next section).
The training corpus has been tokenized using a byte version of [Byte-Pair Encoding (BPE)](https://github.com/openai/gpt-2)
used in the original [RoBERTA](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model with a vocabulary size of 52,000 tokens. The pretraining consists of a masked language model training at the subword level following the approach employed for the RoBERTa base model with the same hyperparameters as in the original work. The training lasted a total of 48 hours with 16 NVIDIA V100 GPUs of 16GB DDRAM, using Adam optimizer with a peak learning rate of 0.0005 and an effective batch size of 2,048 sentences.
## Training corpora and preprocessing
The training corpus is composed of several biomedical corpora in Spanish, collected from publicly available corpora and crawlers.
To obtain a high-quality training corpus, a cleaning pipeline with the following operations has been applied:
- data parsing in different formats
- sentence splitting
- language detection
- filtering of ill-formed sentences
- deduplication of repetitive contents
- keep the original document boundaries
Finally, the corpora are concatenated and further global deduplication among the corpora has been applied.
The result is a medium-size biomedical corpus for Spanish composed of about 963M tokens. The table below shows some basic statistics of the individual cleaned corpora:
| Name | No. tokens | Description |
|-----------------------------------------------------------------------------------------|-------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| [Medical crawler](https://zenodo.org/record/4561970) | 903,558,136 | Crawler of more than 3,000 URLs belonging to Spanish biomedical and health domains. |
| Clinical cases misc. | 102,855,267 | A miscellany of medical content, essentially clinical cases. Note that a clinical case report is a scientific publication where medical practitioners share patient cases and it is different from a clinical note or document. |
| [Scielo](https://zenodo.org/record/2541681#.YlP1DshBwio) | 60,007,289 | Publications written in Spanish crawled from the Spanish SciELO server in 2017. |
| [BARR2_background](https://temu.bsc.es/BARR2/downloads/background_set.raw_text.tar.bz2) | 24,516,442 | Biomedical Abbreviation Recognition and Resolution (BARR2) containing Spanish clinical case study sections from a variety of clinical disciplines. |
| Wikipedia_life_sciences | 13,890,501 | Wikipedia articles crawled 04/01/2021 with the [Wikipedia API python library](https://pypi.org/project/Wikipedia-API/) starting from the "Ciencias\_de\_la\_vida" category up to a maximum of 5 subcategories. Multiple links to the same articles are then discarded to avoid repeating content. |
| Patents | 13,463,387 | Google Patent in Medical Domain for Spain (Spanish). The accepted codes (Medical Domain) for Json files of patents are: "A61B", "A61C","A61F", "A61H", "A61K", "A61L","A61M", "A61B", "A61P". |
| [EMEA](http://opus.nlpl.eu/download.php?f=EMEA/v3/moses/en-es.txt.zip) | 5,377,448 | Spanish-side documents extracted from parallel corpora made out of PDF documents from the European Medicines Agency. |
| [mespen_Medline](https://zenodo.org/record/3562536#.YTt1fH2xXbR) | 4,166,077 | Spanish-side articles extracted from a collection of Spanish-English parallel corpus consisting of biomedical scientific literature. The collection of parallel resources is aggregated from the MedlinePlus source. |
| PubMed | 1,858,966 | Open-access articles from the PubMed repository crawled in 2017. |
## Evaluation and results
The model has been fine-tuned on three Named Entity Recognition (NER) tasks using three clinical NER datasets:
- [PharmaCoNER](https://zenodo.org/record/4270158): is a track on chemical and drug mention recognition from Spanish medical texts (for more info see: https://temu.bsc.es/pharmaconer/).
- [CANTEMIST](https://zenodo.org/record/3978041#.YTt5qH2xXbQ): is a shared task specifically focusing on named entity recognition of tumour morphology, in Spanish (for more info see: https://zenodo.org/record/3978041#.YTt5qH2xXbQ).
- ICTUSnet: consists of 1,006 hospital discharge reports of patients admitted for stroke from 18 different Spanish hospitals. It contains more than 79,000 annotations for 51 different kinds of variables.
We addressed the NER task as a token classification problem using a standard linear layer along with the BIO tagging schema. We compared our models with the general-domain Spanish [roberta-base-bne](https://huggingface.co/PlanTL-GOB-ES/roberta-base-bne), the general-domain multilingual model that supports Spanish [mBERT](https://huggingface.co/bert-base-multilingual-cased), the domain-specific English model [BioBERT](https://huggingface.co/dmis-lab/biobert-base-cased-v1.2), and three domain-specific models based on continual pre-training, [mBERT-Galén](https://ieeexplore.ieee.org/document/9430499), [XLM-R-Galén](https://ieeexplore.ieee.org/document/9430499) and [BETO-Galén](https://ieeexplore.ieee.org/document/9430499).
The table below shows the F1 scores obtained:
| Tasks/Models | bsc-bio-es | XLM-R-Galén | BETO-Galén | mBERT-Galén | mBERT | BioBERT | roberta-base-bne |
|--------------|----------------|--------------------|--------------|--------------|--------------|--------------|------------------|
| PharmaCoNER | **0.8907** | 0.8754 | 0.8537 | 0.8594 | 0.8671 | 0.8545 | 0.8474 |
| CANTEMIST | **0.8220** | 0.8078 | 0.8153 | 0.8168 | 0.8116 | 0.8070 | 0.7875 |
| ICTUSnet | **0.8727** | 0.8716 | 0.8498 | 0.8509 | 0.8631 | 0.8521 | 0.8677 |
The fine-tuning scripts can be found in the official GitHub [repository](https://github.com/PlanTL-GOB-ES/lm-biomedical-clinical-es).
## Intended uses & limitations
The model is ready-to-use only for masked language modelling to perform the Fill Mask task (try the inference API or read the next section)
However, the is intended to be fine-tuned on downstream tasks such as Named Entity Recognition or Text Classification.
## Cite
To be announced soon.
---
## Funding
This work was funded by the Spanish State Secretariat for Digitalization and Artificial Intelligence (SEDIA) within the framework of the Plan-TL.
## Disclaimer
The models published in this repository are intended for a generalist purpose and are available to third parties. These models may have bias and/or any other undesirable distortions.
When third parties, deploy or provide systems and/or services to other parties using any of these models (or using systems based on these models) or become users of the models, they should note that it is their responsibility to mitigate the risks arising from their use and, in any event, to comply with applicable regulations, including regulations regarding the use of artificial intelligence.
In no event shall the owner of the models (SEDIA – State Secretariat for digitalization and artificial intelligence) nor the creator (BSC – Barcelona Supercomputing Center) be liable for any results arising from the use made by third parties of these models.
|
benjamin/gerpt2 | 76b77997c1a715c3cf61a8d086fb75baa3816ded | 2022-05-11T09:17:11.000Z | [
"pytorch",
"tf",
"jax",
"gpt2",
"text-generation",
"de",
"transformers",
"license:mit"
] | text-generation | false | benjamin | null | benjamin/gerpt2 | 904 | 2 | transformers | 1,837 | ---
language: de
widget:
- text: "In einer schockierenden Entdeckung fanden Wissenschaftler eine Herde Einhörner, die in einem abgelegenen, zuvor unerforschten Tal in den Anden lebten."
license: mit
---
# GerPT2
German large and small versions of GPT2:
- https://huggingface.co/benjamin/gerpt2
- https://huggingface.co/benjamin/gerpt2-large
See the [GPT2 model card](https://huggingface.co/gpt2) for considerations on limitations and bias. See the [GPT2 documentation](https://huggingface.co/transformers/model_doc/gpt2.html) for details on GPT2.
## Comparison to [dbmdz/german-gpt2](https://huggingface.co/dbmdz/german-gpt2)
I evaluated both GerPT2-large and the other German GPT2, [dbmdz/german-gpt2](https://huggingface.co/dbmdz/german-gpt2) on the [CC-100](http://data.statmt.org/cc-100/) dataset and on the German Wikipedia:
| | CC-100 (PPL) | Wikipedia (PPL) |
|-------------------|--------------|-----------------|
| dbmdz/german-gpt2 | 49.47 | 62.92 |
| GerPT2 | 24.78 | 35.33 |
| GerPT2-large | __16.08__ | __23.26__ |
| | | |
See the script `evaluate.py` in the [GerPT2 Github repository](https://github.com/bminixhofer/gerpt2) for the code.
## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
tokenizer = AutoTokenizer.from_pretrained("benjamin/gerpt2-large")
model = AutoModelForCausalLM.from_pretrained("benjamin/gerpt2-large")
prompt = "<your prompt>"
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
print(pipe(prompt)[0]["generated_text"])
```
Also, two tricks might improve the generated text:
```python
output = model.generate(
# during training an EOS token was used to mark the beginning of each text
# so it can help to insert it at the start
torch.tensor(
[tokenizer.eos_token_id] + tokenizer.encode(prompt)
).unsqueeze(0),
do_sample=True,
# try setting bad_words_ids=[[0]] to disallow generating an EOS token, without this the model is
# prone to ending generation early because a significant number of texts from the training corpus
# is quite short
bad_words_ids=[[0]],
max_length=max_length,
)[0]
print(tokenizer.decode(output))
```
## Training details
GerPT2-large is trained on the entire German data from the [CC-100 Corpus](http://data.statmt.org/cc-100/) and weights were initialized from the [English GPT2 model](https://huggingface.co/gpt2-large).
GerPT2-large was trained with:
- a batch size of 256
- using OneCycle learning rate with a maximum of 5e-3
- with AdamW with a weight decay of 0.01
- for 2 epochs
Training took roughly 12 days on 8 TPUv3 cores.
To train GerPT2-large, follow these steps. Scripts are located in the [Github repository](https://github.com/bminixhofer/gerpt2):
0. Download and unzip training data from http://data.statmt.org/cc-100/.
1. Train a tokenizer using `prepare/train_tokenizer.py`. As training data for the tokenizer I used a random subset of 5% of the CC-100 data.
2. (optionally) generate a German input embedding matrix with `prepare/generate_aligned_wte.py`. This uses a neat trick to semantically map tokens from the English tokenizer to tokens from the German tokenizer using aligned word embeddings. E. g.:
```
ĠMinde -> Ġleast
Ġjed -> Ġwhatsoever
flughafen -> Air
vermittlung -> employment
teilung -> ignment
ĠInterpretation -> Ġinterpretation
Ġimport -> Ġimported
hansa -> irl
genehmigungen -> exempt
ĠAuflist -> Ġlists
Ġverschwunden -> Ġdisappeared
ĠFlyers -> ĠFlyers
Kanal -> Channel
Ġlehr -> Ġteachers
Ġnahelie -> Ġconvenient
gener -> Generally
mitarbeiter -> staff
```
This helps a lot on a trial run I did, although I wasn't able to do a full comparison due to budget and time constraints. To use this WTE matrix it can be passed via the `wte_path` to the training script. Credit to [this blogpost](https://medium.com/@pierre_guillou/faster-than-training-from-scratch-fine-tuning-the-english-gpt-2-in-any-language-with-hugging-f2ec05c98787) for the idea of initializing GPT2 from English weights.
3. Tokenize the corpus using `prepare/tokenize_text.py`. This generates files for train and validation tokens in JSON Lines format.
4. Run the training script `train.py`! `run.sh` shows how this was executed for the full run with config `configs/tpu_large.json`.
## License
GerPT2 is licensed under the MIT License.
## Citing
Please cite GerPT2 as follows:
```
@misc{Minixhofer_GerPT2_German_large_2020,
author = {Minixhofer, Benjamin},
doi = {10.5281/zenodo.5509984},
month = {12},
title = {{GerPT2: German large and small versions of GPT2}},
url = {https://github.com/bminixhofer/gerpt2},
year = {2020}
}
```
## Acknowledgements
Thanks to [Hugging Face](https://huggingface.co) for awesome tools and infrastructure.
Huge thanks to [Artus Krohn-Grimberghe](https://twitter.com/artuskg) at [LYTiQ](https://www.lytiq.de/) for making this possible by sponsoring the resources used for training. |
JorisCos/ConvTasNet_Libri2Mix_sepnoisy_16k | 98b76c842d1fae9868f74b331b298a92eee3c12e | 2021-09-23T15:48:58.000Z | [
"pytorch",
"dataset:Libri2Mix",
"dataset:sep_noisy",
"asteroid",
"audio",
"ConvTasNet",
"audio-to-audio",
"license:cc-by-sa-4.0"
] | audio-to-audio | false | JorisCos | null | JorisCos/ConvTasNet_Libri2Mix_sepnoisy_16k | 903 | null | asteroid | 1,838 | ---
tags:
- asteroid
- audio
- ConvTasNet
- audio-to-audio
datasets:
- Libri2Mix
- sep_noisy
license: cc-by-sa-4.0
---
## Asteroid model `JorisCos/ConvTasNet_Libri2Mix_sepnoisy_16k`
Description:
This model was trained by Joris Cosentino using the librimix recipe in [Asteroid](https://github.com/asteroid-team/asteroid).
It was trained on the `sep_noisy` task of the Libri2Mix dataset.
Training config:
```yml
data:
n_src: 2
sample_rate: 16000
segment: 3
task: sep_noisy
train_dir: data/wav16k/min/train-360
valid_dir: data/wav16k/min/dev
filterbank:
kernel_size: 32
n_filters: 512
stride: 16
masknet:
bn_chan: 128
hid_chan: 512
mask_act: relu
n_blocks: 8
n_repeats: 3
n_src: 2
skip_chan: 128
optim:
lr: 0.001
optimizer: adam
weight_decay: 0.0
training:
batch_size: 6
early_stop: true
epochs: 200
half_lr: true
num_workers: 4
```
Results:
On Libri2Mix min test set :
```yml
si_sdr: 10.617130949793383
si_sdr_imp: 12.551811412989263
sdr: 11.231867464482065
sdr_imp: 13.059765009747343
sir: 24.461138352988346
sir_imp: 24.371856452307703
sar: 11.5649982725426
sar_imp: 4.662525705768228
stoi: 0.8701085138712695
stoi_imp: 0.2245418019822898
```
License notice:
This work "ConvTasNet_Libri2Mix_sepnoisy_16k" is a derivative of [LibriSpeech ASR corpus](http://www.openslr.org/12) by Vassil Panayotov,
used under[CC BY 4.0](https://creativecommons.org/licenses/by/4.0/); of The WSJ0 Hipster Ambient Mixtures
dataset by [Whisper.ai](http://wham.whisper.ai/), used under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) (Research only).
"ConvTasNet_Libri2Mix_sepnoisy_16k" is licensed under [Attribution-ShareAlike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/) by Joris Cosentino |
facebook/dino-vitb8 | 745a5a92b1e313ab3c2e95a558df5566b5b8e253 | 2021-08-25T17:40:41.000Z | [
"pytorch",
"vit",
"feature-extraction",
"dataset:imagenet-1k",
"arxiv:2010.11929",
"arxiv:2104.14294",
"transformers",
"dino",
"license:apache-2.0"
] | feature-extraction | false | facebook | null | facebook/dino-vitb8 | 903 | 2 | transformers | 1,839 | ---
license: apache-2.0
tags:
- dino
datasets:
- imagenet-1k
---
# Vision Transformer (base-sized model, patch size 8) trained using DINO
Vision Transformer (ViT) model trained using the DINO method. It was introduced in the paper [Emerging Properties in Self-Supervised Vision Transformers](https://arxiv.org/abs/2010.11929) by Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, Armand Joulin and first released in [this repository](https://github.com/facebookresearch/dino).
Disclaimer: The team releasing DINO did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrained on a large collection of images in a self-supervised fashion, namely ImageNet-1k, at a resolution of 224x224 pixels.
Images are presented to the model as a sequence of fixed-size patches (resolution 8x8), which are linearly embedded. One also adds a [CLS] token to the beginning of a sequence to use it for classification tasks. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder.
Note that this model does not include any fine-tuned heads.
By pre-training the model, it learns an inner representation of images that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled images for instance, you can train a standard classifier by placing a linear layer on top of the pre-trained encoder. One typically places a linear layer on top of the [CLS] token, as the last hidden state of this token can be seen as a representation of an entire image.
## Intended uses & limitations
You can use the raw model for image classification. See the [model hub](https://huggingface.co/models?search=google/vit) to look for
fine-tuned versions on a task that interests you.
### How to use
Here is how to use this model:
```python
from transformers import ViTFeatureExtractor, ViTModel
from PIL import Image
import requests
url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)
feature_extractor = ViTFeatureExtractor.from_pretrained('facebook/dino-vitb8')
model = ViTModel.from_pretrained('facebook/dino-vitb8')
inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)
last_hidden_states = outputs.last_hidden_state
```
### BibTeX entry and citation info
```bibtex
@article{DBLP:journals/corr/abs-2104-14294,
author = {Mathilde Caron and
Hugo Touvron and
Ishan Misra and
Herv{\'{e}} J{\'{e}}gou and
Julien Mairal and
Piotr Bojanowski and
Armand Joulin},
title = {Emerging Properties in Self-Supervised Vision Transformers},
journal = {CoRR},
volume = {abs/2104.14294},
year = {2021},
url = {https://arxiv.org/abs/2104.14294},
archivePrefix = {arXiv},
eprint = {2104.14294},
timestamp = {Tue, 04 May 2021 15:12:43 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2104-14294.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
``` |
speechbrain/lang-id-voxlingua107-ecapa | 9835356c3e7d6525f9182813b4b229e9226d53fc | 2022-06-25T03:42:48.000Z | [
"multilingual",
"dataset:VoxLingua107",
"arxiv:2106.04624",
"speechbrain",
"audio-classification",
"embeddings",
"Language",
"Identification",
"pytorch",
"ECAPA-TDNN",
"TDNN",
"VoxLingua107",
"license:apache-2.0"
] | audio-classification | false | speechbrain | null | speechbrain/lang-id-voxlingua107-ecapa | 903 | 6 | speechbrain | 1,840 | ---
language: multilingual
thumbnail:
tags:
- audio-classification
- speechbrain
- embeddings
- Language
- Identification
- pytorch
- ECAPA-TDNN
- TDNN
- VoxLingua107
license: "apache-2.0"
datasets:
- VoxLingua107
metrics:
- Accuracy
widget:
- example_title: English Sample
src: https://cdn-media.huggingface.co/speech_samples/LibriSpeech_61-70968-0000.flac
---
# VoxLingua107 ECAPA-TDNN Spoken Language Identification Model
## Model description
This is a spoken language recognition model trained on the VoxLingua107 dataset using SpeechBrain.
The model uses the ECAPA-TDNN architecture that has previously been used for speaker recognition. However, it uses
more fully connected hidden layers after the embedding layer, and cross-entropy loss was used for training.
We observed that this improved the performance of extracted utterance embeddings for downstream tasks.
The system is trained with recordings sampled at 16kHz (single channel).
The code will automatically normalize your audio (i.e., resampling + mono channel selection) when calling *classify_file* if needed.
The model can classify a speech utterance according to the language spoken.
It covers 107 different languages (
Abkhazian,
Afrikaans,
Amharic,
Arabic,
Assamese,
Azerbaijani,
Bashkir,
Belarusian,
Bulgarian,
Bengali,
Tibetan,
Breton,
Bosnian,
Catalan,
Cebuano,
Czech,
Welsh,
Danish,
German,
Greek,
English,
Esperanto,
Spanish,
Estonian,
Basque,
Persian,
Finnish,
Faroese,
French,
Galician,
Guarani,
Gujarati,
Manx,
Hausa,
Hawaiian,
Hindi,
Croatian,
Haitian,
Hungarian,
Armenian,
Interlingua,
Indonesian,
Icelandic,
Italian,
Hebrew,
Japanese,
Javanese,
Georgian,
Kazakh,
Central Khmer,
Kannada,
Korean,
Latin,
Luxembourgish,
Lingala,
Lao,
Lithuanian,
Latvian,
Malagasy,
Maori,
Macedonian,
Malayalam,
Mongolian,
Marathi,
Malay,
Maltese,
Burmese,
Nepali,
Dutch,
Norwegian Nynorsk,
Norwegian,
Occitan,
Panjabi,
Polish,
Pushto,
Portuguese,
Romanian,
Russian,
Sanskrit,
Scots,
Sindhi,
Sinhala,
Slovak,
Slovenian,
Shona,
Somali,
Albanian,
Serbian,
Sundanese,
Swedish,
Swahili,
Tamil,
Telugu,
Tajik,
Thai,
Turkmen,
Tagalog,
Turkish,
Tatar,
Ukrainian,
Urdu,
Uzbek,
Vietnamese,
Waray,
Yiddish,
Yoruba,
Mandarin Chinese).
## Intended uses & limitations
The model has two uses:
- use 'as is' for spoken language recognition
- use as an utterance-level feature (embedding) extractor, for creating a dedicated language ID model on your own data
The model is trained on automatically collected YouTube data. For more
information about the dataset, see [here](http://bark.phon.ioc.ee/voxlingua107/).
#### How to use
```python
import torchaudio
from speechbrain.pretrained import EncoderClassifier
language_id = EncoderClassifier.from_hparams(source="speechbrain/lang-id-voxlingua107-ecapa", savedir="tmp")
# Download Thai language sample from Omniglot and cvert to suitable form
signal = language_id.load_audio("https://omniglot.com/soundfiles/udhr/udhr_th.mp3")
prediction = language_id.classify_batch(signal)
print(prediction)
# (tensor([[-2.8646e+01, -3.0346e+01, -2.0748e+01, -2.9562e+01, -2.2187e+01,
# -3.2668e+01, -3.6677e+01, -3.3573e+01, -3.2545e+01, -2.4365e+01,
# -2.4688e+01, -3.1171e+01, -2.7743e+01, -2.9918e+01, -2.4770e+01,
# -3.2250e+01, -2.4727e+01, -2.6087e+01, -2.1870e+01, -3.2821e+01,
# -2.2128e+01, -2.2822e+01, -3.0888e+01, -3.3564e+01, -2.9906e+01,
# -2.2392e+01, -2.5573e+01, -2.6443e+01, -3.2429e+01, -3.2652e+01,
# -3.0030e+01, -2.4607e+01, -2.2967e+01, -2.4396e+01, -2.8578e+01,
# -2.5153e+01, -2.8475e+01, -2.6409e+01, -2.5230e+01, -2.7957e+01,
# -2.6298e+01, -2.3609e+01, -2.5863e+01, -2.8225e+01, -2.7225e+01,
# -3.0486e+01, -2.1185e+01, -2.7938e+01, -3.3155e+01, -1.9076e+01,
# -2.9181e+01, -2.2160e+01, -1.8352e+01, -2.5866e+01, -3.3636e+01,
# -4.2016e+00, -3.1581e+01, -3.1894e+01, -2.7834e+01, -2.5429e+01,
# -3.2235e+01, -3.2280e+01, -2.8786e+01, -2.3366e+01, -2.6047e+01,
# -2.2075e+01, -2.3770e+01, -2.2518e+01, -2.8101e+01, -2.5745e+01,
# -2.6441e+01, -2.9822e+01, -2.7109e+01, -3.0225e+01, -2.4566e+01,
# -2.9268e+01, -2.7651e+01, -3.4221e+01, -2.9026e+01, -2.6009e+01,
# -3.1968e+01, -3.1747e+01, -2.8156e+01, -2.9025e+01, -2.7756e+01,
# -2.8052e+01, -2.9341e+01, -2.8806e+01, -2.1636e+01, -2.3992e+01,
# -2.3794e+01, -3.3743e+01, -2.8332e+01, -2.7465e+01, -1.5085e-02,
# -2.9094e+01, -2.1444e+01, -2.9780e+01, -3.6046e+01, -3.7401e+01,
# -3.0888e+01, -3.3172e+01, -1.8931e+01, -2.2679e+01, -3.0225e+01,
# -2.4995e+01, -2.1028e+01]]), tensor([-0.0151]), tensor([94]), ['th'])
# The scores in the prediction[0] tensor can be interpreted as log-likelihoods that
# the given utterance belongs to the given language (i.e., the larger the better)
# The linear-scale likelihood can be retrieved using the following:
print(prediction[1].exp())
# tensor([0.9850])
# The identified language ISO code is given in prediction[3]
print(prediction[3])
# ['th: Thai']
# Alternatively, use the utterance embedding extractor:
emb = language_id.encode_batch(signal)
print(emb.shape)
# torch.Size([1, 1, 256])
```
To perform inference on the GPU, add `run_opts={"device":"cuda"}` when calling the `from_hparams` method.
The system is trained with recordings sampled at 16kHz (single channel).
The code will automatically normalize your audio (i.e., resampling + mono channel selection) when calling *classify_file* if needed. Make sure your input tensor is compliant with the expected sampling rate if you use *encode_batch* and *classify_batch*.
#### Limitations and bias
Since the model is trained on VoxLingua107, it has many limitations and biases, some of which are:
- Probably it's accuracy on smaller languages is quite limited
- Probably it works worse on female speech than male speech (because YouTube data includes much more male speech)
- Based on subjective experiments, it doesn't work well on speech with a foreign accent
- Probably it doesn't work well on children's speech and on persons with speech disorders
## Training data
The model is trained on [VoxLingua107](http://bark.phon.ioc.ee/voxlingua107/).
VoxLingua107 is a speech dataset for training spoken language identification models.
The dataset consists of short speech segments automatically extracted from YouTube videos and labeled according the language of the video title and description, with some post-processing steps to filter out false positives.
VoxLingua107 contains data for 107 languages. The total amount of speech in the training set is 6628 hours.
The average amount of data per language is 62 hours. However, the real amount per language varies a lot. There is also a seperate development set containing 1609 speech segments from 33 languages, validated by at least two volunteers to really contain the given language.
## Training procedure
See the [SpeechBrain recipe](https://github.com/speechbrain/speechbrain/tree/voxlingua107/recipes/VoxLingua107/lang_id).
## Evaluation results
Error rate: 6.7% on the VoxLingua107 development dataset
#### Referencing SpeechBrain
```bibtex
@misc{speechbrain,
title={{SpeechBrain}: A General-Purpose Speech Toolkit},
author={Mirco Ravanelli and Titouan Parcollet and Peter Plantinga and Aku Rouhe and Samuele Cornell and Loren Lugosch and Cem Subakan and Nauman Dawalatabad and Abdelwahab Heba and Jianyuan Zhong and Ju-Chieh Chou and Sung-Lin Yeh and Szu-Wei Fu and Chien-Feng Liao and Elena Rastorgueva and François Grondin and William Aris and Hwidong Na and Yan Gao and Renato De Mori and Yoshua Bengio},
year={2021},
eprint={2106.04624},
archivePrefix={arXiv},
primaryClass={eess.AS},
note={arXiv:2106.04624}
}
```
### Referencing VoxLingua107
```bibtex
@inproceedings{valk2021slt,
title={{VoxLingua107}: a Dataset for Spoken Language Recognition},
author={J{\"o}rgen Valk and Tanel Alum{\"a}e},
booktitle={Proc. IEEE SLT Workshop},
year={2021},
}
```
#### About SpeechBrain
SpeechBrain is an open-source and all-in-one speech toolkit. It is designed to be simple, extremely flexible, and user-friendly. Competitive or state-of-the-art performance is obtained in various domains.
Website: https://speechbrain.github.io/
GitHub: https://github.com/speechbrain/speechbrain
|
navteca/bart-large-mnli | c39c03bcf29d1dab341409eee0b8cd3d7fa68b8a | 2021-08-06T13:59:01.000Z | [
"pytorch",
"jax",
"bart",
"text-classification",
"en",
"dataset:multi_nli",
"arxiv:1909.00161",
"transformers",
"zero-shot-classification",
"license:mit"
] | zero-shot-classification | false | navteca | null | navteca/bart-large-mnli | 902 | 2 | transformers | 1,841 | ---
datasets:
- multi_nli
language: en
license: mit
pipeline_tag: zero-shot-classification
tags:
- bart
- zero-shot-classification
---
# Bart large model for NLI-based Zero Shot Text Classification
This model uses [bart-large](https://huggingface.co/facebook/bart-large).
## Training Data
This model was trained on the [MultiNLI (MNLI)](https://huggingface.co/datasets/multi_nli) dataset in the manner originally described in [Yin et al. 2019](https://arxiv.org/abs/1909.00161).
It can be used to predict whether a topic label can be assigned to a given sequence, whether or not the label has been seen before.
## Usage and Performance
The trained model can be used like this:
```python
from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
# Load model & tokenizer
bart_model = AutoModelForSequenceClassification.from_pretrained('navteca/bart-large-mnli')
bart_tokenizer = AutoTokenizer.from_pretrained('navteca/bart-large-mnli')
# Get predictions
nlp = pipeline('zero-shot-classification', model=bart_model, tokenizer=bart_tokenizer)
sequence = 'One day I will see the world.'
candidate_labels = ['cooking', 'dancing', 'travel']
result = nlp(sequence, candidate_labels, multi_label=True)
print(result)
#{
# "sequence": "One day I will see the world.",
# "labels": [
# "travel",
# "dancing",
# "cooking"
# ],
# "scores": [
# 0.9941897988319397,
# 0.0060537424869835,
# 0.0020010927692056
# ]
#}
``` |
CAMeL-Lab/bert-base-arabic-camelbert-mix | 9be352797bdf28a9ae21e2ae582aaaca7abdb22d | 2021-09-14T14:34:32.000Z | [
"pytorch",
"tf",
"jax",
"bert",
"fill-mask",
"ar",
"arxiv:2103.06678",
"transformers",
"Arabic",
"Dialect",
"Egyptian",
"Gulf",
"Levantine",
"Classical Arabic",
"MSA",
"Modern Standard Arabic",
"license:apache-2.0",
"autotrain_compatible"
] | fill-mask | false | CAMeL-Lab | null | CAMeL-Lab/bert-base-arabic-camelbert-mix | 901 | 6 | transformers | 1,842 | ---
language:
- ar
license: apache-2.0
tags:
- Arabic
- Dialect
- Egyptian
- Gulf
- Levantine
- Classical Arabic
- MSA
- Modern Standard Arabic
widget:
- text: "الهدف من الحياة هو [MASK] ."
---
# CAMeLBERT: A collection of pre-trained models for Arabic NLP tasks
## Model description
**CAMeLBERT** is a collection of BERT models pre-trained on Arabic texts with different sizes and variants.
We release pre-trained language models for Modern Standard Arabic (MSA), dialectal Arabic (DA), and classical Arabic (CA), in addition to a model pre-trained on a mix of the three.
We also provide additional models that are pre-trained on a scaled-down set of the MSA variant (half, quarter, eighth, and sixteenth).
The details are described in the paper *"[The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models](https://arxiv.org/abs/2103.06678)."*
This model card describes **CAMeLBERT-Mix** (`bert-base-arabic-camelbert-mix`), a model pre-trained on a mixture of these variants: MSA, DA, and CA.
||Model|Variant|Size|#Word|
|-|-|:-:|-:|-:|
|✔|`bert-base-arabic-camelbert-mix`|CA,DA,MSA|167GB|17.3B|
||`bert-base-arabic-camelbert-ca`|CA|6GB|847M|
||`bert-base-arabic-camelbert-da`|DA|54GB|5.8B|
||`bert-base-arabic-camelbert-msa`|MSA|107GB|12.6B|
||`bert-base-arabic-camelbert-msa-half`|MSA|53GB|6.3B|
||`bert-base-arabic-camelbert-msa-quarter`|MSA|27GB|3.1B|
||`bert-base-arabic-camelbert-msa-eighth`|MSA|14GB|1.6B|
||`bert-base-arabic-camelbert-msa-sixteenth`|MSA|6GB|746M|
## Intended uses
You can use the released model for either masked language modeling or next sentence prediction.
However, it is mostly intended to be fine-tuned on an NLP task, such as NER, POS tagging, sentiment analysis, dialect identification, and poetry classification.
We release our fine-tuninig code [here](https://github.com/CAMeL-Lab/CAMeLBERT).
#### How to use
You can use this model directly with a pipeline for masked language modeling:
```python
>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='CAMeL-Lab/bert-base-arabic-camelbert-mix')
>>> unmasker("الهدف من الحياة هو [MASK] .")
[{'sequence': '[CLS] الهدف من الحياة هو النجاح. [SEP]',
'score': 0.10861027985811234,
'token': 6232,
'token_str': 'النجاح'},
{'sequence': '[CLS] الهدف من الحياة هو.. [SEP]',
'score': 0.07626965641975403,
'token': 18,
'token_str': '.'},
{'sequence': '[CLS] الهدف من الحياة هو الحياة. [SEP]',
'score': 0.05131986364722252,
'token': 3696,
'token_str': 'الحياة'},
{'sequence': '[CLS] الهدف من الحياة هو الموت. [SEP]',
'score': 0.03734956309199333,
'token': 4295,
'token_str': 'الموت'},
{'sequence': '[CLS] الهدف من الحياة هو العمل. [SEP]',
'score': 0.027189988642930984,
'token': 2854,
'token_str': 'العمل'}]
```
*Note*: to download our models, you would need `transformers>=3.5.0`. Otherwise, you could download the models manually.
Here is how to use this model to get the features of a given text in PyTorch:
```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained('CAMeL-Lab/bert-base-arabic-camelbert-mix')
model = AutoModel.from_pretrained('CAMeL-Lab/bert-base-arabic-camelbert-mix')
text = "مرحبا يا عالم."
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
```
and in TensorFlow:
```python
from transformers import AutoTokenizer, TFAutoModel
tokenizer = AutoTokenizer.from_pretrained('CAMeL-Lab/bert-base-arabic-camelbert-mix')
model = TFAutoModel.from_pretrained('CAMeL-Lab/bert-base-arabic-camelbert-mix')
text = "مرحبا يا عالم."
encoded_input = tokenizer(text, return_tensors='tf')
output = model(encoded_input)
```
## Training data
- MSA (Modern Standard Arabic)
- [The Arabic Gigaword Fifth Edition](https://catalog.ldc.upenn.edu/LDC2011T11)
- [Abu El-Khair Corpus](http://www.abuelkhair.net/index.php/en/arabic/abu-el-khair-corpus)
- [OSIAN corpus](https://vlo.clarin.eu/search;jsessionid=31066390B2C9E8C6304845BA79869AC1?1&q=osian)
- [Arabic Wikipedia](https://archive.org/details/arwiki-20190201)
- The unshuffled version of the Arabic [OSCAR corpus](https://oscar-corpus.com/)
- DA (dialectal Arabic)
- A collection of dialectal Arabic data described in [our paper](https://arxiv.org/abs/2103.06678).
- CA (classical Arabic)
- [OpenITI (Version 2020.1.2)](https://zenodo.org/record/3891466#.YEX4-F0zbzc)
## Training procedure
We use [the original implementation](https://github.com/google-research/bert) released by Google for pre-training.
We follow the original English BERT model's hyperparameters for pre-training, unless otherwise specified.
### Preprocessing
- After extracting the raw text from each corpus, we apply the following pre-processing.
- We first remove invalid characters and normalize white spaces using the utilities provided by [the original BERT implementation](https://github.com/google-research/bert/blob/eedf5716ce1268e56f0a50264a88cafad334ac61/tokenization.py#L286-L297).
- We also remove lines without any Arabic characters.
- We then remove diacritics and kashida using [CAMeL Tools](https://github.com/CAMeL-Lab/camel_tools).
- Finally, we split each line into sentences with a heuristics-based sentence segmenter.
- We train a WordPiece tokenizer on the entire dataset (167 GB text) with a vocabulary size of 30,000 using [HuggingFace's tokenizers](https://github.com/huggingface/tokenizers).
- We do not lowercase letters nor strip accents.
### Pre-training
- The model was trained on a single cloud TPU (`v3-8`) for one million steps in total.
- The first 90,000 steps were trained with a batch size of 1,024 and the rest was trained with a batch size of 256.
- The sequence length was limited to 128 tokens for 90% of the steps and 512 for the remaining 10%.
- We use whole word masking and a duplicate factor of 10.
- We set max predictions per sequence to 20 for the dataset with max sequence length of 128 tokens and 80 for the dataset with max sequence length of 512 tokens.
- We use a random seed of 12345, masked language model probability of 0.15, and short sequence probability of 0.1.
- The optimizer used is Adam with a learning rate of 1e-4, \\(\beta_{1} = 0.9\\) and \\(\beta_{2} = 0.999\\), a weight decay of 0.01, learning rate warmup for 10,000 steps and linear decay of the learning rate after.
## Evaluation results
- We evaluate our pre-trained language models on five NLP tasks: NER, POS tagging, sentiment analysis, dialect identification, and poetry classification.
- We fine-tune and evaluate the models using 12 dataset.
- We used Hugging Face's transformers to fine-tune our CAMeLBERT models.
- We used transformers `v3.1.0` along with PyTorch `v1.5.1`.
- The fine-tuning was done by adding a fully connected linear layer to the last hidden state.
- We use \\(F_{1}\\) score as a metric for all tasks.
- Code used for fine-tuning is available [here](https://github.com/CAMeL-Lab/CAMeLBERT).
### Results
| Task | Dataset | Variant | Mix | CA | DA | MSA | MSA-1/2 | MSA-1/4 | MSA-1/8 | MSA-1/16 |
| -------------------- | --------------- | ------- | ----- | ----- | ----- | ----- | ------- | ------- | ------- | -------- |
| NER | ANERcorp | MSA | 80.8% | 67.9% | 74.1% | 82.4% | 82.0% | 82.1% | 82.6% | 80.8% |
| POS | PATB (MSA) | MSA | 98.1% | 97.8% | 97.7% | 98.3% | 98.2% | 98.3% | 98.2% | 98.2% |
| | ARZTB (EGY) | DA | 93.6% | 92.3% | 92.7% | 93.6% | 93.6% | 93.7% | 93.6% | 93.6% |
| | Gumar (GLF) | DA | 97.3% | 97.7% | 97.9% | 97.9% | 97.9% | 97.9% | 97.9% | 97.9% |
| SA | ASTD | MSA | 76.3% | 69.4% | 74.6% | 76.9% | 76.0% | 76.8% | 76.7% | 75.3% |
| | ArSAS | MSA | 92.7% | 89.4% | 91.8% | 93.0% | 92.6% | 92.5% | 92.5% | 92.3% |
| | SemEval | MSA | 69.0% | 58.5% | 68.4% | 72.1% | 70.7% | 72.8% | 71.6% | 71.2% |
| DID | MADAR-26 | DA | 62.9% | 61.9% | 61.8% | 62.6% | 62.0% | 62.8% | 62.0% | 62.2% |
| | MADAR-6 | DA | 92.5% | 91.5% | 92.2% | 91.9% | 91.8% | 92.2% | 92.1% | 92.0% |
| | MADAR-Twitter-5 | MSA | 75.7% | 71.4% | 74.2% | 77.6% | 78.5% | 77.3% | 77.7% | 76.2% |
| | NADI | DA | 24.7% | 17.3% | 20.1% | 24.9% | 24.6% | 24.6% | 24.9% | 23.8% |
| Poetry | APCD | CA | 79.8% | 80.9% | 79.6% | 79.7% | 79.9% | 80.0% | 79.7% | 79.8% |
### Results (Average)
| | Variant | Mix | CA | DA | MSA | MSA-1/2 | MSA-1/4 | MSA-1/8 | MSA-1/16 |
| -------------------- | ------- | ----- | ----- | ----- | ----- | ------- | ------- | ------- | -------- |
| Variant-wise-average<sup>[[1]](#footnote-1)</sup> | MSA | 82.1% | 75.7% | 80.1% | 83.4% | 83.0% | 83.3% | 83.2% | 82.3% |
| | DA | 74.4% | 72.1% | 72.9% | 74.2% | 74.0% | 74.3% | 74.1% | 73.9% |
| | CA | 79.8% | 80.9% | 79.6% | 79.7% | 79.9% | 80.0% | 79.7% | 79.8% |
| Macro-Average | ALL | 78.7% | 74.7% | 77.1% | 79.2% | 79.0% | 79.2% | 79.1% | 78.6% |
<a name="footnote-1">[1]</a>: Variant-wise-average refers to average over a group of tasks in the same language variant.
## Acknowledgements
This research was supported with Cloud TPUs from Google’s TensorFlow Research Cloud (TFRC).
## Citation
```bibtex
@inproceedings{inoue-etal-2021-interplay,
title = "The Interplay of Variant, Size, and Task Type in {A}rabic Pre-trained Language Models",
author = "Inoue, Go and
Alhafni, Bashar and
Baimukan, Nurpeiis and
Bouamor, Houda and
Habash, Nizar",
booktitle = "Proceedings of the Sixth Arabic Natural Language Processing Workshop",
month = apr,
year = "2021",
address = "Kyiv, Ukraine (Online)",
publisher = "Association for Computational Linguistics",
abstract = "In this paper, we explore the effects of language variants, data sizes, and fine-tuning task types in Arabic pre-trained language models. To do so, we build three pre-trained language models across three variants of Arabic: Modern Standard Arabic (MSA), dialectal Arabic, and classical Arabic, in addition to a fourth language model which is pre-trained on a mix of the three. We also examine the importance of pre-training data size by building additional models that are pre-trained on a scaled-down set of the MSA variant. We compare our different models to each other, as well as to eight publicly available models by fine-tuning them on five NLP tasks spanning 12 datasets. Our results suggest that the variant proximity of pre-training data to fine-tuning data is more important than the pre-training data size. We exploit this insight in defining an optimized system selection model for the studied tasks.",
}
```
|
castorini/ance-msmarco-doc-maxp | 95207533d035adaddeb195da8484cb6cfaa366f3 | 2021-05-20T15:17:50.000Z | [
"pytorch",
"roberta",
"arxiv:2007.00808",
"transformers"
] | null | false | castorini | null | castorini/ance-msmarco-doc-maxp | 901 | null | transformers | 1,843 | This model is converted from the original ANCE [repo](https://github.com/microsoft/ANCE) and fitted into Pyserini:
> Lee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang, Jialin Liu, Paul Bennett, Junaid Ahmed, Arnold Overwijk. [Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval](https://arxiv.org/pdf/2007.00808.pdf)
For more details on how to use it, check our experiments in [Pyserini](https://github.com/castorini/pyserini/blob/master/docs/experiments-ance.md)
|
M-CLIP/M-BERT-Base-69 | e5bf2855224ca5294be65b45344ddcd06219c41d | 2021-05-18T21:33:14.000Z | [
"pytorch",
"jax",
"bert",
"feature-extraction",
"transformers"
] | feature-extraction | false | M-CLIP | null | M-CLIP/M-BERT-Base-69 | 900 | null | transformers | 1,844 | <br />
<p align="center">
<h1 align="center">M-BERT Base 69</h1>
<p align="center">
<a href="https://github.com/FreddeFrallan/Multilingual-CLIP/tree/main/Model%20Cards/M-BERT%20Base%2069">Github Model Card</a>
</p>
</p>
## Usage
To use this model along with the original CLIP vision encoder you need to download the code and additional linear weights from the [Multilingual-CLIP Github](https://github.com/FreddeFrallan/Multilingual-CLIP).
Once this is done, you can load and use the model with the following code
```python
from src import multilingual_clip
model = multilingual_clip.load_model('M-BERT-Base-40')
embeddings = model(['Älgen är skogens konung!', 'Wie leben Eisbären in der Antarktis?', 'Вы знали, что все белые медведи левши?'])
print(embeddings.shape)
# Yields: torch.Size([3, 640])
```
<!-- ABOUT THE PROJECT -->
## About
A [BERT-base-multilingual](https://huggingface.co/bert-base-multilingual-cased) tuned to match the embedding space for [69 languages](https://github.com/FreddeFrallan/Multilingual-CLIP/blob/main/Model%20Cards/M-BERT%20Base%2069/Fine-Tune-Languages.md), to the embedding space of the CLIP text encoder which accompanies the Res50x4 vision encoder. <br>
A full list of the 100 languages used during pre-training can be found [here](https://github.com/google-research/bert/blob/master/multilingual.md#list-of-languages), and a list of the 4069languages used during fine-tuning can be found in [SupportedLanguages.md](https://github.com/FreddeFrallan/Multilingual-CLIP/blob/main/Model%20Cards/M-BERT%20Base%2069/Fine-Tune-Languages.md).
Training data pairs was generated by sampling 40k sentences for each language from the combined descriptions of [GCC](https://ai.google.com/research/ConceptualCaptions/) + [MSCOCO](https://cocodataset.org/#home) + [VizWiz](https://vizwiz.org/tasks-and-datasets/image-captioning/), and translating them into the corresponding language.
All translation was done using the [AWS translate service](https://aws.amazon.com/translate/), the quality of these translations have currently not been analyzed, but one can assume the quality varies between the 69 languages.
|
SkolkovoInstitute/rubert-base-corruption-detector | 27965caf27a4897bd0df76128dc8707ca7e212a7 | 2021-12-18T09:28:50.000Z | [
"pytorch",
"bert",
"text-classification",
"ru",
"transformers",
"fluency"
] | text-classification | false | SkolkovoInstitute | null | SkolkovoInstitute/rubert-base-corruption-detector | 900 | null | transformers | 1,845 | ---
language:
- ru
tags:
- fluency
---
This is a model for evaluation of naturalness of short Russian texts. It has been trained to distinguish human-written texts from their corrupted versions.
Corruption sources: random replacement, deletion, addition, shuffling, and re-inflection of words and characters, random changes of capitalization, round-trip translation, filling random gaps with T5 and RoBERTA models. For each original text, we sampled three corrupted texts, so the model is uniformly biased towards the `unnatural` label.
Data sources: web-corpora from [the Leipzig collection](https://wortschatz.uni-leipzig.de/en/download) (`rus_news_2020_100K`, `rus_newscrawl-public_2018_100K`, `rus-ru_web-public_2019_100K`, `rus_wikipedia_2021_100K`), comments from [OK](https://www.kaggle.com/alexandersemiletov/toxic-russian-comments) and [Pikabu](https://www.kaggle.com/blackmoon/russian-language-toxic-comments).
On our private test dataset, the model has achieved 40% rank correlation with human judgements of naturalness, which is higher than GPT perplexity, another popular fluency metric. |
monologg/koelectra-base-v2-finetuned-korquad | 6cf15019cbd304a9cef33a1c94e66850814a66f9 | 2020-06-03T03:32:20.000Z | [
"pytorch",
"electra",
"question-answering",
"transformers",
"autotrain_compatible"
] | question-answering | false | monologg | null | monologg/koelectra-base-v2-finetuned-korquad | 900 | null | transformers | 1,846 | Entry not found |
juliamendelsohn/framing_issue_generic | 6b35ca7630b0b6fb208e600ed4f3c236f8abe042 | 2021-05-20T17:27:30.000Z | [
"pytorch",
"roberta",
"transformers"
] | null | false | juliamendelsohn | null | juliamendelsohn/framing_issue_generic | 898 | null | transformers | 1,847 | Entry not found |
facebook/detr-resnet-101-panoptic | 4297151a5e287d2ed673392d2c0a1c6d46142d5c | 2022-06-27T08:34:50.000Z | [
"pytorch",
"detr",
"image-segmentation",
"dataset:coco",
"arxiv:2005.12872",
"transformers",
"vision",
"license:apache-2.0"
] | image-segmentation | false | facebook | null | facebook/detr-resnet-101-panoptic | 897 | 2 | transformers | 1,848 | ---
license: apache-2.0
tags:
- image-segmentation
- vision
datasets:
- coco
widget:
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/dog-cat.jpg
example_title: Dog & Cat
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/construction-site.jpg
example_title: Construction Site
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/apple-orange.jpg
example_title: Apple & Orange
---
# DETR (End-to-End Object Detection) model with ResNet-101 backbone
DEtection TRansformer (DETR) model trained end-to-end on COCO 2017 panoptic (118k annotated images). It was introduced in the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Carion et al. and first released in [this repository](https://github.com/facebookresearch/detr).
Disclaimer: The team releasing DETR did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
The DETR model is an encoder-decoder transformer with a convolutional backbone. Two heads are added on top of the decoder outputs in order to perform object detection: a linear layer for the class labels and a MLP (multi-layer perceptron) for the bounding boxes. The model uses so-called object queries to detect objects in an image. Each object query looks for a particular object in the image. For COCO, the number of object queries is set to 100.
The model is trained using a "bipartite matching loss": one compares the predicted classes + bounding boxes of each of the N = 100 object queries to the ground truth annotations, padded up to the same length N (so if an image only contains 4 objects, 96 annotations will just have a "no object" as class and "no bounding box" as bounding box). The Hungarian matching algorithm is used to create an optimal one-to-one mapping between each of the N queries and each of the N annotations. Next, standard cross-entropy (for the classes) and a linear combination of the L1 and generalized IoU loss (for the bounding boxes) are used to optimize the parameters of the model.
DETR can be naturally extended to perform panoptic segmentation, by adding a mask head on top of the decoder outputs.
## Intended uses & limitations
You can use the raw model for panoptic segmentation. See the [model hub](https://huggingface.co/models?search=facebook/detr) to look for all available DETR models.
### How to use
Here is how to use this model:
```python
from transformers import DetrFeatureExtractor, DetrForSegmentation
from PIL import Image
import requests
url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)
feature_extractor = DetrFeatureExtractor.from_pretrained('facebook/detr-resnet-101-panoptic')
model = DetrForSegmentation.from_pretrained('facebook/detr-resnet-101-panoptic')
inputs = feature_extractor(images=image, return_tensors="pt")
outputs = model(**inputs)
# model predicts COCO classes, bounding boxes, and masks
logits = outputs.logits
bboxes = outputs.pred_boxes
masks = outputs.pred_masks
```
Currently, both the feature extractor and model support PyTorch.
## Training data
The DETR model was trained on [COCO 2017 panoptic](https://cocodataset.org/#download), a dataset consisting of 118k/5k annotated images for training/validation respectively.
## Training procedure
### Preprocessing
The exact details of preprocessing of images during training/validation can be found [here](https://github.com/facebookresearch/detr/blob/master/datasets/coco_panoptic.py).
Images are resized/rescaled such that the shortest side is at least 800 pixels and the largest side at most 1333 pixels, and normalized across the RGB channels with the ImageNet mean (0.485, 0.456, 0.406) and standard deviation (0.229, 0.224, 0.225).
### Training
The model was trained for 300 epochs on 16 V100 GPUs. This takes 3 days, with 4 images per GPU (hence a total batch size of 64).
## Evaluation results
This model achieves the following results on COCO 2017 validation: a box AP (average precision) of **40.1**, a segmentation AP (average precision) of **33** and a PQ (panoptic quality) of **45.1**.
For more details regarding evaluation results, we refer to table 5 of the original paper.
### BibTeX entry and citation info
```bibtex
@article{DBLP:journals/corr/abs-2005-12872,
author = {Nicolas Carion and
Francisco Massa and
Gabriel Synnaeve and
Nicolas Usunier and
Alexander Kirillov and
Sergey Zagoruyko},
title = {End-to-End Object Detection with Transformers},
journal = {CoRR},
volume = {abs/2005.12872},
year = {2020},
url = {https://arxiv.org/abs/2005.12872},
archivePrefix = {arXiv},
eprint = {2005.12872},
timestamp = {Thu, 28 May 2020 17:38:09 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2005-12872.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
``` |
allenai/PRIMERA-multixscience | 69ef13b16f5edc76323f57af922c9b4c47bb7d5c | 2022-07-25T18:17:07.000Z | [
"pytorch",
"led",
"text2text-generation",
"transformers",
"license:apache-2.0",
"autotrain_compatible"
] | text2text-generation | false | allenai | null | allenai/PRIMERA-multixscience | 896 | 1 | transformers | 1,849 | ---
license: apache-2.0
---
HF-version model for PRIMERA: Pyramid-based Masked Sentence Pre-training for Multi-document Summarization (ACL 2022).
The original code can be found [here](https://github.com/allenai/PRIMER). You can find the script and notebook to train/evaluate the model in the original github repo.
* Note: due to the difference between the implementations of the original Longformer and the Huggingface LED model, the results of converted models are slightly different. We run a sanity check on both fine-tuned and non fine-tuned models on the **MultiNews dataset**, and show the results below:
| Model | Rouge-1 | Rouge-2 | Rouge-L |
| --- | ----------- |----------- |----------- |
| PRIMERA | 42.0 | 13.6 | 20.8|
| PRIMERA-hf | 41.7 |13.6 | 20.5|
| PRIMERA(finetuned) | 49.9 | 21.1 | 25.9|
| PRIMERA-hf(finetuned) | 49.9 | 20.9 | 25.8|
You can use it by
```
from transformers import (
AutoTokenizer,
LEDConfig,
LEDForConditionalGeneration,
)
tokenizer = AutoTokenizer.from_pretrained('allenai/PRIMERA')
config=LEDConfig.from_pretrained('allenai/PRIMERA')
model = LEDForConditionalGeneration.from_pretrained('allenai/PRIMERA')
``` |
tner/roberta-large-tweetner-2020 | 768bf9f64587af66884cfb5053999594f341baa9 | 2022-07-08T11:45:18.000Z | [
"pytorch",
"roberta",
"token-classification",
"transformers",
"autotrain_compatible"
] | token-classification | false | tner | null | tner/roberta-large-tweetner-2020 | 896 | null | transformers | 1,850 | Entry not found |
ThomasNLG/t5-qa_squad2neg-en | 41de3e39d518801383740526946e70880d096cd8 | 2021-07-09T07:44:39.000Z | [
"pytorch",
"jax",
"t5",
"text2text-generation",
"en",
"dataset:squad_v2",
"arxiv:2103.12693",
"transformers",
"qa",
"question",
"answering",
"SQuAD",
"metric",
"nlg",
"t5-small",
"license:mit",
"model-index",
"autotrain_compatible"
] | text2text-generation | false | ThomasNLG | null | ThomasNLG/t5-qa_squad2neg-en | 895 | null | transformers | 1,851 | ---
language: en
tags:
- qa
- question
- answering
- SQuAD
- metric
- nlg
- t5-small
license: mit
datasets:
- squad_v2
model-index:
- name: t5-qa_squad2neg-en
results:
- task:
name: Question Answering
type: extractive-qa
widget:
- text: "Who was Louis 14? </s> Louis 14 was a French King."
---
# t5-qa_squad2neg-en
## Model description
This model is a *Question Answering* model based on T5-small.
It is actually a component of [QuestEval](https://github.com/ThomasScialom/QuestEval) metric but can be used independently as it is, for QA only.
## How to use
```python
from transformers import T5Tokenizer, T5ForConditionalGeneration
tokenizer = T5Tokenizer.from_pretrained("ThomasNLG/t5-qa_squad2neg-en")
model = T5ForConditionalGeneration.from_pretrained("ThomasNLG/t5-qa_squad2neg-en")
```
You can play with the model using the inference API, the text input format should follow this template (accordingly to the training stage of the model):
`text_input = "{QUESTION} </s> {CONTEXT}"`
## Training data
The model was trained on:
- SQuAD-v2
- SQuAD-v2 neg: in addition to the training data of SQuAD-v2, for each answerable example, a negative sampled example has been added with the label *unanswerable* to help the model learning when the question is not answerable given the context. For more details, see the [paper](https://arxiv.org/abs/2103.12693).
### Citation info
```bibtex
@article{scialom2020QuestEval,
title={QuestEval: Summarization Asks for Fact-based Evaluation},
author={Scialom, Thomas and Dray, Paul-Alexis and Gallinari, Patrick and Lamprier, Sylvain and Piwowarski, Benjamin and Staiano, Jacopo and Wang, Alex},
journal={arXiv preprint arXiv:2103.12693},
year={2021}
}
```
|
OFA-Sys/OFA-base | 01ecca4855f318a69ed4821957ee23d499d28cc3 | 2022-07-25T11:52:55.000Z | [
"pytorch",
"ofa",
"transformers",
"license:apache-2.0"
] | null | false | OFA-Sys | null | OFA-Sys/OFA-base | 893 | 2 | transformers | 1,852 | ---
license: apache-2.0
---
# OFA-base
This is the **base** version of OFA pretrained model. OFA is a unified multimodal pretrained model that unifies modalities (i.e., cross-modality, vision, language) and tasks (e.g., image generation, visual grounding, image captioning, image classification, text generation, etc.) to a simple sequence-to-sequence learning framework.
The directory includes 4 files, namely `config.json` which consists of model configuration, `vocab.json` and `merge.txt` for our OFA tokenizer, and lastly `pytorch_model.bin` which consists of model weights. There is no need to worry about the mismatch between Fairseq and transformers, since we have addressed the issue yet.
To use it in transformers, please refer to https://github.com/OFA-Sys/OFA/tree/feature/add_transformers. Install the transformers and download the models as shown below.
```
git clone --single-branch --branch feature/add_transformers https://github.com/OFA-Sys/OFA.git
pip install OFA/transformers/
git clone https://huggingface.co/OFA-Sys/OFA-base
```
After, refer the path to OFA-base to `ckpt_dir`, and prepare an image for the testing example below. Also, ensure that you have pillow and torchvision in your environment.
```
>>> from PIL import Image
>>> from torchvision import transforms
>>> from transformers import OFATokenizer, OFAModel
>>> from generate import sequence_generator
>>> mean, std = [0.5, 0.5, 0.5], [0.5, 0.5, 0.5]
>>> resolution = 384
>>> patch_resize_transform = transforms.Compose([
lambda image: image.convert("RGB"),
transforms.Resize((resolution, resolution), interpolation=Image.BICUBIC),
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std)
])
>>> tokenizer = OFATokenizer.from_pretrained(ckpt_dir)
>>> txt = " what does the image describe?"
>>> inputs = tokenizer([txt], return_tensors="pt").input_ids
>>> img = Image.open(path_to_image)
>>> patch_img = patch_resize_transform(img).unsqueeze(0)
>>> # using the generator of fairseq version
>>> model = OFAModel.from_pretrained(ckpt_dir, use_cache=True)
>>> generator = sequence_generator.SequenceGenerator(
tokenizer=tokenizer,
beam_size=5,
max_len_b=16,
min_len=0,
no_repeat_ngram_size=3,
)
>>> data = {}
>>> data["net_input"] = {"input_ids": inputs, 'patch_images': patch_img, 'patch_masks':torch.tensor([True])}
>>> gen_output = generator.generate([model], data)
>>> gen = [gen_output[i][0]["tokens"] for i in range(len(gen_output))]
>>> # using the generator of huggingface version
>>> model = OFAModel.from_pretrained(ckpt_dir, use_cache=False)
>>> gen = model.generate(inputs, patch_images=patch_img, num_beams=5, no_repeat_ngram_size=3)
>>> print(tokenizer.batch_decode(gen, skip_special_tokens=True))
```
|
asahi417/relbert-roberta-large | 07c0062eb062303e48c0fe2544148af5fd6c76a4 | 2021-07-05T13:39:36.000Z | [
"pytorch",
"roberta",
"feature-extraction",
"transformers"
] | feature-extraction | false | asahi417 | null | asahi417/relbert-roberta-large | 892 | null | transformers | 1,853 | # RelBERT
RoBERTa finetuned on the contrastive loss for lexical relation. Please take a look [the official repository](https://github.com/asahi417/relbert).
|
barissayil/bert-sentiment-analysis-sst | 969d390abf6b567c74ce1af74505b449734b4285 | 2021-06-11T09:47:14.000Z | [
"pytorch",
"bert",
"text-classification",
"transformers"
] | text-classification | false | barissayil | null | barissayil/bert-sentiment-analysis-sst | 891 | null | transformers | 1,854 | Entry not found |
MaryaAI/opus-mt-en-ar-finetuned-Math-13-10-en-to-ar | 133d8b58a906746884d78764aeb064bad1871dae | 2021-10-17T08:27:27.000Z | [
"pytorch",
"tensorboard",
"marian",
"text2text-generation",
"dataset:syssr_en_ar",
"transformers",
"generated_from_trainer",
"license:apache-2.0",
"model-index",
"autotrain_compatible"
] | text2text-generation | false | MaryaAI | null | MaryaAI/opus-mt-en-ar-finetuned-Math-13-10-en-to-ar | 890 | null | transformers | 1,855 | ---
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- syssr_en_ar
model-index:
- name: opus-mt-en-ar-finetuned-Math-13-10-en-to-ar
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. -->
# opus-mt-en-ar-finetuned-Math-13-10-en-to-ar
This model is a fine-tuned version of [Helsinki-NLP/opus-mt-en-ar](https://huggingface.co/Helsinki-NLP/opus-mt-en-ar) on the syssr_en_ar dataset.
## 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: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5
- mixed_precision_training: Native AMP
### Framework versions
- Transformers 4.11.3
- Pytorch 1.9.0+cu111
- Datasets 1.13.0
- Tokenizers 0.10.3
|
textattack/distilbert-base-uncased-imdb | 5b0f46c2fc4b86bf21f0ec0409bed77ee142b332 | 2020-07-06T16:34:50.000Z | [
"pytorch",
"distilbert",
"text-classification",
"transformers"
] | text-classification | false | textattack | null | textattack/distilbert-base-uncased-imdb | 889 | null | transformers | 1,856 | ## TextAttack Model Card
This `distilbert-base-uncased` model was fine-tuned for sequence classification using TextAttack
and the imdb dataset loaded using the `nlp` library. The model was fine-tuned
for 5 epochs with a batch size of 16, a learning
rate of 2e-05, and a maximum sequence length of 128.
Since this was a classification task, the model was trained with a cross-entropy loss function.
The best score the model achieved on this task was 0.88, as measured by the
eval set accuracy, found after 2 epochs.
For more information, check out [TextAttack on Github](https://github.com/QData/TextAttack).
|
tanmoyio/wav2vec2-large-xlsr-bengali | 7447c623dca066b74d0299d0132dfca0674b6c8a | 2021-09-23T16:39:27.000Z | [
"pytorch",
"wav2vec2",
"automatic-speech-recognition",
"Bengali",
"dataset:OpenSLR",
"transformers",
"audio",
"speech",
"xlsr-fine-tuning-week",
"license:cc-by-sa-4.0",
"model-index"
] | automatic-speech-recognition | false | tanmoyio | null | tanmoyio/wav2vec2-large-xlsr-bengali | 888 | 2 | transformers | 1,857 | ---
language: Bengali
datasets:
- OpenSLR
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: cc-by-sa-4.0
model-index:
- name: XLSR Wav2Vec2 Bengali by Tanmoy Sarkar
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: OpenSLR
type: OpenSLR
args: ben
metrics:
- name: Test WER
type: wer
value: 88.58
---
# Wav2Vec2-Large-XLSR-Bengali
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) Bengali using the [Bengali ASR training data set containing ~196K utterances](https://www.openslr.org/53/).
When using this model, make sure that your speech input is sampled at 16kHz.
## Usage
Dataset must be downloaded from [this website](https://www.openslr.org/53/) and preprocessed accordingly. For example 1250 test samples has been chosen.
```python
import pandas as pd
test_dataset = pd.read_csv('utt_spk_text.tsv', sep='\\t', header=None)[60000:61250]
test_dataset.columns = ["audio_path", "__", "label"]
test_dataset = test_data.drop("__", axis=1)
def add_file_path(text):
path = "data/" + text[:2] + "/" + text + '.flac'
return path
test_dataset['audio_path'] = test_dataset['audio_path'].map(lambda x: add_file_path(x))
```
The model can be used directly (without a language model) as follows:
```python
import torch
import torchaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
processor = Wav2Vec2Processor.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
model = Wav2Vec2ForCTC.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["audio_path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["label"][:2])
```
## Evaluation
The model can be evaluated as follows on the Bengali test data of OpenSLR.
```python
import torch
import torchaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
model = Wav2Vec2ForCTC.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
model.to("cuda")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["label"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = processor.batch_decode(pred_ids)
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```
**Test Result**: 88.58 %
## Training
The script used for training can be found [Bengali ASR Fine Tuning Wav2Vec2](https://colab.research.google.com/drive/1Bkc5C_cJV9BeS0FD0MuHyayl8hqcbdRZ?usp=sharing)
|
Helsinki-NLP/opus-mt-tc-big-en-es | 26b349a9177b11b92b2b56b47344ebe73e515817 | 2022-06-01T12:59:20.000Z | [
"pytorch",
"marian",
"text2text-generation",
"en",
"es",
"transformers",
"translation",
"opus-mt-tc",
"license:cc-by-4.0",
"model-index",
"autotrain_compatible"
] | translation | false | Helsinki-NLP | null | Helsinki-NLP/opus-mt-tc-big-en-es | 887 | null | transformers | 1,858 | ---
language:
- en
- es
tags:
- translation
- opus-mt-tc
license: cc-by-4.0
model-index:
- name: opus-mt-tc-big-en-es
results:
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: flores101-devtest
type: flores_101
args: eng spa devtest
metrics:
- name: BLEU
type: bleu
value: 28.5
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: news-test2008
type: news-test2008
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 30.1
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: tatoeba-test-v2021-08-07
type: tatoeba_mt
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 57.2
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: tico19-test
type: tico19-test
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 53.0
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: newstest2009
type: wmt-2009-news
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 30.2
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: newstest2010
type: wmt-2010-news
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 37.6
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: newstest2011
type: wmt-2011-news
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 38.9
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: newstest2012
type: wmt-2012-news
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 39.5
- task:
name: Translation eng-spa
type: translation
args: eng-spa
dataset:
name: newstest2013
type: wmt-2013-news
args: eng-spa
metrics:
- name: BLEU
type: bleu
value: 35.9
---
# opus-mt-tc-big-en-es
Neural machine translation model for translating from English (en) to Spanish (es).
This model is part of the [OPUS-MT project](https://github.com/Helsinki-NLP/Opus-MT), an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of [Marian NMT](https://marian-nmt.github.io/), an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from [OPUS](https://opus.nlpl.eu/) and training pipelines use the procedures of [OPUS-MT-train](https://github.com/Helsinki-NLP/Opus-MT-train).
* Publications: [OPUS-MT – Building open translation services for the World](https://aclanthology.org/2020.eamt-1.61/) and [The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/) (Please, cite if you use this model.)
```
@inproceedings{tiedemann-thottingal-2020-opus,
title = "{OPUS}-{MT} {--} Building open translation services for the World",
author = {Tiedemann, J{\"o}rg and Thottingal, Santhosh},
booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
month = nov,
year = "2020",
address = "Lisboa, Portugal",
publisher = "European Association for Machine Translation",
url = "https://aclanthology.org/2020.eamt-1.61",
pages = "479--480",
}
@inproceedings{tiedemann-2020-tatoeba,
title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
author = {Tiedemann, J{\"o}rg},
booktitle = "Proceedings of the Fifth Conference on Machine Translation",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2020.wmt-1.139",
pages = "1174--1182",
}
```
## Model info
* Release: 2022-03-13
* source language(s): eng
* target language(s): spa
* model: transformer-big
* data: opusTCv20210807+bt ([source](https://github.com/Helsinki-NLP/Tatoeba-Challenge))
* tokenization: SentencePiece (spm32k,spm32k)
* original model: [opusTCv20210807+bt_transformer-big_2022-03-13.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-spa/opusTCv20210807+bt_transformer-big_2022-03-13.zip)
* more information released models: [OPUS-MT eng-spa README](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-spa/README.md)
## Usage
A short example code:
```python
from transformers import MarianMTModel, MarianTokenizer
src_text = [
"A wasp stung him and he had an allergic reaction.",
"I love nature."
]
model_name = "pytorch-models/opus-mt-tc-big-en-es"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
for t in translated:
print( tokenizer.decode(t, skip_special_tokens=True) )
# expected output:
# Una avispa lo picó y tuvo una reacción alérgica.
# Me encanta la naturaleza.
```
You can also use OPUS-MT models with the transformers pipelines, for example:
```python
from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-en-es")
print(pipe("A wasp stung him and he had an allergic reaction."))
# expected output: Una avispa lo picó y tuvo una reacción alérgica.
```
## Benchmarks
* test set translations: [opusTCv20210807+bt_transformer-big_2022-03-13.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-spa/opusTCv20210807+bt_transformer-big_2022-03-13.test.txt)
* test set scores: [opusTCv20210807+bt_transformer-big_2022-03-13.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-spa/opusTCv20210807+bt_transformer-big_2022-03-13.eval.txt)
* benchmark results: [benchmark_results.txt](benchmark_results.txt)
* benchmark output: [benchmark_translations.zip](benchmark_translations.zip)
| langpair | testset | chr-F | BLEU | #sent | #words |
|----------|---------|-------|-------|-------|--------|
| eng-spa | tatoeba-test-v2021-08-07 | 0.73863 | 57.2 | 16583 | 134710 |
| eng-spa | flores101-devtest | 0.56440 | 28.5 | 1012 | 29199 |
| eng-spa | newssyscomb2009 | 0.58415 | 31.5 | 502 | 12503 |
| eng-spa | news-test2008 | 0.56707 | 30.1 | 2051 | 52586 |
| eng-spa | newstest2009 | 0.57836 | 30.2 | 2525 | 68111 |
| eng-spa | newstest2010 | 0.62357 | 37.6 | 2489 | 65480 |
| eng-spa | newstest2011 | 0.62415 | 38.9 | 3003 | 79476 |
| eng-spa | newstest2012 | 0.63031 | 39.5 | 3003 | 79006 |
| eng-spa | newstest2013 | 0.60354 | 35.9 | 3000 | 70528 |
| eng-spa | tico19-test | 0.73554 | 53.0 | 2100 | 66563 |
## Acknowledgements
The work is supported by the [European Language Grid](https://www.european-language-grid.eu/) as [pilot project 2866](https://live.european-language-grid.eu/catalogue/#/resource/projects/2866), by the [FoTran project](https://www.helsinki.fi/en/researchgroups/natural-language-understanding-with-cross-lingual-grounding), funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 771113), and the [MeMAD project](https://memad.eu/), funded by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by [CSC -- IT Center for Science](https://www.csc.fi/), Finland.
## Model conversion info
* transformers version: 4.16.2
* OPUS-MT git hash: 3405783
* port time: Wed Apr 13 18:03:53 EEST 2022
* port machine: LM0-400-22516.local
|
shibing624/mengzi-t5-base-chinese-correction | 091e91da1215be5f40ae8d2273a7fe0b93b5354f | 2022-06-17T08:23:49.000Z | [
"pytorch",
"t5",
"text2text-generation",
"zh",
"transformers",
"license:apache-2.0",
"autotrain_compatible"
] | text2text-generation | false | shibing624 | null | shibing624/mengzi-t5-base-chinese-correction | 887 | 2 | transformers | 1,859 | ---
language:
- zh
tags:
- t5
- pytorch
- zh
license: "apache-2.0"
---
# T5 for Chinese Spelling Correction Model
中文拼写纠错模型
`shibing624/mengzi-t5-base-chinese-correction` evaluate SIGHAN2015 test data：
- Sentence Level: precision:0.8321, recall:0.6390, f1:0.7229
训练使用的数据集为下方提供的“SIGHAN+Wang271K中文纠错数据集”，在SIGHAN2015的测试集上达到接近SOTA水平。
未改动模型结构，finetune中文纠错数据集，评估纠错效果很好，模型潜力巨大。
## Usage
本项目开源在中文文本纠错项目：[pycorrector](https://github.com/shibing624/pycorrector)，可支持t5模型，通过如下命令调用：
```
pip install -U pycorrector
```
run:
```python
from pycorrector.t5.t5_corrector import T5Corrector
nlp = T5Corrector("shibing624/mengzi-t5-base-chinese-correction").batch_t5_correct
i = "今天新情很好"
print(i, ' => ', nlp([i]))
```
output:
```shell
今天新情很好 => 今天心情很好 [('新', '心', 2, 3)]
```
模型文件组成：
```
mengzi-t5-base-chinese-correction
|-- config.json
|-- pytorch_model.bin
|-- special_tokens_map.json
|-- spiece.model
|-- tokenizer_config.json
`-- tokenizer.json
```
如果需要训练t5-correction，请参考[https://github.com/shibing624/pycorrector/tree/master/pycorrector/t5](https://github.com/shibing624/pycorrector/tree/master/pycorrector/t5)
### 训练数据集
#### SIGHAN+Wang271K中文纠错数据集
| 数据集 | 语料 | 下载链接 | 压缩包大小 |
| :------- | :--------- | :---------: | :---------: |
| **`SIGHAN+Wang271K中文纠错数据集`** | SIGHAN+Wang271K(27万条) | [百度网盘（密码01b9）](https://pan.baidu.com/s/1BV5tr9eONZCI0wERFvr0gQ)| 106M |
| **`原始SIGHAN数据集`** | SIGHAN13 14 15 | [官方csc.html](http://nlp.ee.ncu.edu.tw/resource/csc.html)| 339K |
| **`原始Wang271K数据集`** | Wang271K | [Automatic-Corpus-Generation dimmywang提供](https://github.com/wdimmy/Automatic-Corpus-Generation/blob/master/corpus/train.sgml)| 93M |
SIGHAN+Wang271K中文纠错数据集，数据格式：
```json
[
{
"id": "B2-4029-3",
"original_text": "晚间会听到嗓音，白天的时候大家都不会太在意，但是在睡觉的时候这嗓音成为大家的恶梦。",
"wrong_ids": [
5,
31
],
"correct_text": "晚间会听到噪音，白天的时候大家都不会太在意，但是在睡觉的时候这噪音成为大家的恶梦。"
},
]
```
## Citation
```latex
@software{pycorrector,
author = {Xu Ming},
title = {pycorrector: Text Error Correction Tool},
year = {2021},
url = {https://github.com/shibing624/pycorrector},
}
```
|
moha/arabert_c19 | eab96c316448fe535686332e35be64949f6ab7d7 | 2021-05-19T23:35:40.000Z | [
"pytorch",
"jax",
"bert",
"fill-mask",
"ar",
"arxiv:2105.03143",
"arxiv:2004.04315",
"transformers",
"autotrain_compatible"
] | fill-mask | false | moha | null | moha/arabert_c19 | 885 | null | transformers | 1,860 | ---
language: ar
widget:
- text: "لكي نتجنب فيروس [MASK]"
---
# arabert_c19: An Arabert model pretrained on 1.5 million COVID-19 multi-dialect Arabic tweets
**ARABERT COVID-19** [Arxiv URL](https://arxiv.org/pdf/2105.03143.pdf) is a pretrained (fine-tuned) version of the AraBERT v2 model (https://huggingface.co/aubmindlab/bert-base-arabertv02). The pretraining was done using 1.5 million multi-dialect Arabic tweets regarding the COVID-19 pandemic from the “Large Arabic Twitter Dataset on COVID-19” (https://arxiv.org/abs/2004.04315).
The model can achieve better results for the tasks that deal with multi-dialect Arabic tweets in relation to the COVID-19 pandemic.
# Classification results for multiple tasks including fake-news and hate speech detection when using arabert_c19 and mbert_ar_c19:
For more details refer to the paper (link)
| | arabert | mbert | distilbert multi | arabert Covid-19 | mbert Covid-19 |
|------------------------------------|----------|----------|------------------|------------------|----------------|
| Contains hate (Binary) | 0.8346 | 0.6675 | 0.7145 | `0.8649` | 0.8492 |
| Talk about a cure (Binary) | 0.8193 | 0.7406 | 0.7127 | 0.9055 | `0.9176` |
| News or opinion (Binary) | 0.8987 | 0.8332 | 0.8099 | `0.9163` | 0.9116 |
| Contains fake information (Binary) | 0.6415 | 0.5428 | 0.4743 | `0.7739` | 0.7228 |
# Preprocessing
```python
from arabert.preprocess import ArabertPreprocessor
model_name="moha/arabert_c19"
arabert_prep = ArabertPreprocessor(model_name=model_name)
text = "للوقايه من عدم انتشار كورونا عليك اولا غسل اليدين بالماء والصابون وتكون عملية الغسل دقيقه تشمل راحة اليد الأصابع التركيز على الإبهام"
arabert_prep.preprocess(text)
```
# Citation
Please cite as:
``` bibtex
@misc{ameur2021aracovid19mfh,
title={AraCOVID19-MFH: Arabic COVID-19 Multi-label Fake News and Hate Speech Detection Dataset},
author={Mohamed Seghir Hadj Ameur and Hassina Aliane},
year={2021},
eprint={2105.03143},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
# Contacts
**Hadj Ameur**: [Github](https://github.com/MohamedHadjAmeur) | <[email protected]> | <[email protected]>
|
PlanTL-GOB-ES/roberta-base-biomedical-es | d672dee3226a354e0f9e5c11369fad6b6cb1f522 | 2022-04-08T14:10:27.000Z | [
"pytorch",
"roberta",
"fill-mask",
"es",
"arxiv:2109.03570",
"arxiv:2109.07765",
"transformers",
"biomedical",
"spanish",
"license:apache-2.0",
"autotrain_compatible"
] | fill-mask | false | PlanTL-GOB-ES | null | PlanTL-GOB-ES/roberta-base-biomedical-es | 884 | 1 | transformers | 1,861 | ---
language:
- es
tags:
- biomedical
- spanish
license: apache-2.0
metrics:
- ppl
widget:
- text: "El único antecedente personal a reseñar era la <mask> arterial."
- text: "Las radiologías óseas de cuerpo entero no detectan alteraciones <mask>, ni alteraciones vertebrales."
- text: "En el <mask> toraco-abdómino-pélvico no se encontraron hallazgos patológicos de interés."
---
# Biomedical language model for Spanish
Biomedical pretrained language model for Spanish. For more details about the corpus, the pretraining and the evaluation, check the official [repository](https://github.com/PlanTL-SANIDAD/lm-biomedical-clinical-es) and read our [preprint](https://arxiv.org/abs/2109.03570) "_Carrino, C. P., Armengol-Estapé, J., Gutiérrez-Fandiño, A., Llop-Palao, J., Pàmies, M., Gonzalez-Agirre, A., & Villegas, M. (2021). Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario._".
## Tokenization and model pretraining
This model is a [RoBERTa-based](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model trained on a
**biomedical** corpus in Spanish collected from several sources (see next section).
The training corpus has been tokenized using a byte version of [Byte-Pair Encoding (BPE)](https://github.com/openai/gpt-2)
used in the original [RoBERTA](https://github.com/pytorch/fairseq/tree/master/examples/roberta) model with a vocabulary size of 52,000 tokens. The pretraining consists of a masked language model training at the subword level following the approach employed for the RoBERTa base model with the same hyperparameters as in the original work. The training lasted a total of 48 hours with 16 NVIDIA V100 GPUs of 16GB DDRAM, using Adam optimizer with a peak learning rate of 0.0005 and an effective batch size of 2,048 sentences.
## Training corpora and preprocessing
The training corpus is composed of several biomedical corpora in Spanish, collected from publicly available corpora and crawlers.
To obtain a high-quality training corpus, a cleaning pipeline with the following operations has been applied:
- data parsing in different formats
- sentence splitting
- language detection
- filtering of ill-formed sentences
- deduplication of repetitive contents
- keep the original document boundaries
Finally, the corpora are concatenated and further global deduplication among the corpora have been applied.
The result is a medium-size biomedical corpus for Spanish composed of about 963M tokens. The table below shows some basic statistics of the individual cleaned corpora:
| Name | No. tokens | Description |
|-----------------------------------------------------------------------------------------|-------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| [Medical crawler](https://zenodo.org/record/4561970) | 745,705,946 | Crawler of more than 3,000 URLs belonging to Spanish biomedical and health domains. |
| Clinical cases misc. | 102,855,267 | A miscellany of medical content, essentially clinical cases. Note that a clinical case report is a scientific publication where medical practitioners share patient cases and it is different from a clinical note or document. |
| [Scielo](https://github.com/PlanTL-SANIDAD/SciELO-Spain-Crawler) | 60,007,289 | Publications written in Spanish crawled from the Spanish SciELO server in 2017. |
| [BARR2_background](https://temu.bsc.es/BARR2/downloads/background_set.raw_text.tar.bz2) | 24,516,442 | Biomedical Abbreviation Recognition and Resolution (BARR2) containing Spanish clinical case study sections from a variety of clinical disciplines. |
| Wikipedia_life_sciences | 13,890,501 | Wikipedia articles crawled 04/01/2021 with the [Wikipedia API python library](https://pypi.org/project/Wikipedia-API/) starting from the "Ciencias\_de\_la\_vida" category up to a maximum of 5 subcategories. Multiple links to the same articles are then discarded to avoid repeating content. |
| Patents | 13,463,387 | Google Patent in Medical Domain for Spain (Spanish). The accepted codes (Medical Domain) for Json files of patents are: "A61B", "A61C","A61F", "A61H", "A61K", "A61L","A61M", "A61B", "A61P". |
| [EMEA](http://opus.nlpl.eu/download.php?f=EMEA/v3/moses/en-es.txt.zip) | 5,377,448 | Spanish-side documents extracted from parallel corpora made out of PDF documents from the European Medicines Agency. |
| [mespen_Medline](https://zenodo.org/record/3562536#.YTt1fH2xXbR) | 4,166,077 | Spanish-side articles extracted from a collection of Spanish-English parallel corpus consisting of biomedical scientific literature. The collection of parallel resources are aggregated from the MedlinePlus source. |
| PubMed | 1,858,966 | Open-access articles from the PubMed repository crawled in 2017. |
## Evaluation and results
The model has been evaluated on the Named Entity Recognition (NER) using the following datasets:
- [PharmaCoNER](https://zenodo.org/record/4270158): is a track on chemical and drug mention recognition from Spanish medical texts (for more info see: https://temu.bsc.es/pharmaconer/).
- [CANTEMIST](https://zenodo.org/record/3978041#.YTt5qH2xXbQ): is a shared task specifically focusing on named entity recognition of tumor morphology, in Spanish (for more info see: https://zenodo.org/record/3978041#.YTt5qH2xXbQ).
- ICTUSnet: consists of 1,006 hospital discharge reports of patients admitted for stroke from 18 different Spanish hospitals. It contains more than 79,000 annotations for 51 different kinds of variables.
The evaluation results are compared against the [mBERT](https://huggingface.co/bert-base-multilingual-cased) and [BETO](https://huggingface.co/dccuchile/bert-base-spanish-wwm-cased) models:
| F1 - Precision - Recall | roberta-base-biomedical-es | mBERT | BETO |
|---------------------------|----------------------------|-------------------------------|-------------------------|
| PharmaCoNER | **89.48** - **87.85** - **91.18** | 87.46 - 86.50 - 88.46 | 88.18 - 87.12 - 89.28 |
| CANTEMIST | **83.87** - **81.70** - **86.17** | 82.61 - 81.12 - 84.15 | 82.42 - 80.91 - 84.00 |
| ICTUSnet | **88.12** - **85.56** - **90.83** | 86.75 - 83.53 - 90.23 | 85.95 - 83.10 - 89.02 |
## Intended uses & limitations
The model is ready-to-use only for masked language modelling to perform the Fill Mask task (try the inference API or read the next section)
However, the is intended to be fine-tuned on downstream tasks such as Named Entity Recognition or Text Classification.
## Cite
If you use our models, please cite our latest preprint:
```bibtex
@misc{carrino2021biomedical,
title={Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario},
author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Asier Gutiérrez-Fandiño and Joan Llop-Palao and Marc Pàmies and Aitor Gonzalez-Agirre and Marta Villegas},
year={2021},
eprint={2109.03570},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
If you use our Medical Crawler corpus, please cite the preprint:
```bibtex
@misc{carrino2021spanish,
title={Spanish Biomedical Crawled Corpus: A Large, Diverse Dataset for Spanish Biomedical Language Models},
author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Ona de Gibert Bonet and Asier Gutiérrez-Fandiño and Aitor Gonzalez-Agirre and Martin Krallinger and Marta Villegas},
year={2021},
eprint={2109.07765},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
---
## How to use
```python
from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")
model = AutoModelForMaskedLM.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")
from transformers import pipeline
unmasker = pipeline('fill-mask', model="BSC-TeMU/roberta-base-biomedical-es")
unmasker("El único antecedente personal a reseñar era la <mask> arterial.")
```
```
# Output
[
{
"sequence": " El único antecedente personal a reseñar era la hipertensión arterial.",
"score": 0.9855039715766907,
"token": 3529,
"token_str": " hipertensión"
},
{
"sequence": " El único antecedente personal a reseñar era la diabetes arterial.",
"score": 0.0039140828885138035,
"token": 1945,
"token_str": " diabetes"
},
{
"sequence": " El único antecedente personal a reseñar era la hipotensión arterial.",
"score": 0.002484665485098958,
"token": 11483,
"token_str": " hipotensión"
},
{
"sequence": " El único antecedente personal a reseñar era la Hipertensión arterial.",
"score": 0.0023484621196985245,
"token": 12238,
"token_str": " Hipertensión"
},
{
"sequence": " El único antecedente personal a reseñar era la presión arterial.",
"score": 0.0008009297889657319,
"token": 2267,
"token_str": " presión"
}
]
``` |
twigs/bart-text2text-simplifier | f2071131fa949dda7dbc24734c13443efaa51da3 | 2022-07-18T21:21:06.000Z | [
"pytorch",
"bart",
"text2text-generation",
"transformers",
"autotrain_compatible"
] | text2text-generation | false | twigs | null | twigs/bart-text2text-simplifier | 884 | null | transformers | 1,862 | Entry not found |
johngiorgi/declutr-sci-base | 34a174c9f34455c1f2705060742d46785ee2de02 | 2022-03-11T14:47:33.000Z | [
"pytorch",
"jax",
"bert",
"fill-mask",
"arxiv:2006.03659",
"transformers",
"autotrain_compatible"
] | fill-mask | false | johngiorgi | null | johngiorgi/declutr-sci-base | 882 | 3 | transformers | 1,863 | # DeCLUTR-sci-base
## Model description
This is the [allenai/scibert_scivocab_uncased](https://huggingface.co/allenai/scibert_scivocab_uncased) model, with extended pretraining on over 2 million scientific papers from [S2ORC](https://github.com/allenai/s2orc/) using the self-supervised training strategy presented in [DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations](https://arxiv.org/abs/2006.03659).
## Intended uses & limitations
The model is intended to be used as a sentence encoder, similar to [Google's Universal Sentence Encoder](https://tfhub.dev/google/universal-sentence-encoder/4) or [Sentence Transformers](https://github.com/UKPLab/sentence-transformers). It is particularly suitable for scientific text.
#### How to use
Please see [our repo](https://github.com/JohnGiorgi/DeCLUTR) for full details. A simple example is shown below.
##### With [SentenceTransformers](https://www.sbert.net/)
```python
from scipy.spatial.distance import cosine
from sentence_transformers import SentenceTransformer
# Load the model
model = SentenceTransformer("johngiorgi/declutr-sci-base")
# Prepare some text to embed
text = [
"Oncogenic KRAS mutations are common in cancer.",
"Notably, c-Raf has recently been found essential for development of K-Ras-driven NSCLCs.",
]
# Embed the text
embeddings = model.encode(texts)
# Compute a semantic similarity via the cosine distance
semantic_sim = 1 - cosine(embeddings[0], embeddings[1])
```
##### With 🤗 Transformers
```python
import torch
from scipy.spatial.distance import cosine
from transformers import AutoModel, AutoTokenizer
# Load the model
tokenizer = AutoTokenizer.from_pretrained("johngiorgi/declutr-sci-base")
model = AutoModel.from_pretrained("johngiorgi/declutr-sci-base")
# Prepare some text to embed
text = [
"Oncogenic KRAS mutations are common in cancer.",
"Notably, c-Raf has recently been found essential for development of K-Ras-driven NSCLCs.",
]
inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt")
# Embed the text
with torch.no_grad():
sequence_output = model(**inputs)[0]
# Mean pool the token-level embeddings to get sentence-level embeddings
embeddings = torch.sum(
sequence_output * inputs["attention_mask"].unsqueeze(-1), dim=1
) / torch.clamp(torch.sum(inputs["attention_mask"], dim=1, keepdims=True), min=1e-9)
# Compute a semantic similarity via the cosine distance
semantic_sim = 1 - cosine(embeddings[0], embeddings[1])
```
### BibTeX entry and citation info
```bibtex
@article{Giorgi2020DeCLUTRDC,
title={DeCLUTR: Deep Contrastive Learning for Unsupervised Textual Representations},
author={John M Giorgi and Osvald Nitski and Gary D. Bader and Bo Wang},
journal={ArXiv},
year={2020},
volume={abs/2006.03659}
}
``` |
Artem1/grammar_error_correcter_v1 | 376d3055b766ba13f6fc92152d6b5a25dd2f11e7 | 2022-07-14T18:54:45.000Z | [
"pytorch",
"t5",
"text2text-generation",
"transformers",
"autotrain_compatible"
] | text2text-generation | false | Artem1 | null | Artem1/grammar_error_correcter_v1 | 882 | null | transformers | 1,864 | Entry not found |
hfl/chinese-electra-base-discriminator | 44c5a47c42df39b11e9841ed602b2d49cdddd1af | 2021-03-03T01:40:07.000Z | [
"pytorch",
"tf",
"electra",
"zh",
"arxiv:2004.13922",
"transformers",
"license:apache-2.0"
] | null | false | hfl | null | hfl/chinese-electra-base-discriminator | 879 | 2 | transformers | 1,865 | ---
language:
- zh
license: "apache-2.0"
---
**Please use `ElectraForPreTraining` for `discriminator` and `ElectraForMaskedLM` for `generator` if you are re-training these models.**
## Chinese ELECTRA
Google and Stanford University released a new pre-trained model called ELECTRA, which has a much compact model size and relatively competitive performance compared to BERT and its variants.
For further accelerating the research of the Chinese pre-trained model, the Joint Laboratory of HIT and iFLYTEK Research (HFL) has released the Chinese ELECTRA models based on the official code of ELECTRA.
ELECTRA-small could reach similar or even higher scores on several NLP tasks with only 1/10 parameters compared to BERT and its variants.
This project is based on the official code of ELECTRA: [https://github.com/google-research/electra](https://github.com/google-research/electra)
You may also interested in,
- Chinese BERT series: https://github.com/ymcui/Chinese-BERT-wwm
- Chinese ELECTRA: https://github.com/ymcui/Chinese-ELECTRA
- Chinese XLNet: https://github.com/ymcui/Chinese-XLNet
- Knowledge Distillation Toolkit - TextBrewer: https://github.com/airaria/TextBrewer
More resources by HFL: https://github.com/ymcui/HFL-Anthology
## Citation
If you find our resource or paper is useful, please consider including the following citation in your paper.
- https://arxiv.org/abs/2004.13922
```
@inproceedings{cui-etal-2020-revisiting,
title = "Revisiting Pre-Trained Models for {C}hinese Natural Language Processing",
author = "Cui, Yiming and
Che, Wanxiang and
Liu, Ting and
Qin, Bing and
Wang, Shijin and
Hu, Guoping",
booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings",
month = nov,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/2020.findings-emnlp.58",
pages = "657--668",
}
```
|
monologg/koelectra-base-v3-hate-speech | 8938df1530df593b9fce6926d1ff963ad07d23a3 | 2020-12-31T12:56:18.000Z | [
"pytorch",
"electra",
"text-classification",
"transformers"
] | text-classification | false | monologg | null | monologg/koelectra-base-v3-hate-speech | 876 | 2 | transformers | 1,866 | Entry not found |
fnlp/cpt-large | d3df73c7677993da4e871d9fcb1239469e52127c | 2022-07-18T08:01:01.000Z | [
"pytorch",
"bart",
"feature-extraction",
"zh",
"arxiv:2109.05729",
"transformers",
"fill-mask",
"text2text-generation",
"text-classification",
"Summarization",
"Chinese",
"CPT",
"BART",
"BERT",
"seq2seq"
] | text-classification | false | fnlp | null | fnlp/cpt-large | 875 | 7 | transformers | 1,867 | ---
tags:
- fill-mask
- text2text-generation
- fill-mask
- text-classification
- Summarization
- Chinese
- CPT
- BART
- BERT
- seq2seq
language: zh
---
# Chinese CPT-Large
## Model description
This is an implementation of CPT-Large. To use CPT, please import the file `modeling_cpt.py` (**Download** [Here](https://github.com/fastnlp/CPT/blob/master/finetune/modeling_cpt.py)) that define the architecture of CPT into your project.
[**CPT: A Pre-Trained Unbalanced Transformer for Both Chinese Language Understanding and Generation**](https://arxiv.org/pdf/2109.05729.pdf)
Yunfan Shao, Zhichao Geng, Yitao Liu, Junqi Dai, Fei Yang, Li Zhe, Hujun Bao, Xipeng Qiu
**Github Link:** https://github.com/fastnlp/CPT
## Usage
```python
>>> from modeling_cpt import CPTForConditionalGeneration
>>> from transformers import BertTokenizer
>>> tokenizer = BertTokenizer.from_pretrained("fnlp/cpt-large")
>>> model = CPTForConditionalGeneration.from_pretrained("fnlp/cpt-large")
>>> input_ids = tokenizer.encode("北京是[MASK]的首都", return_tensors='pt')
>>> pred_ids = model.generate(input_ids, num_beams=4, max_length=20)
>>> print(tokenizer.convert_ids_to_tokens(pred_ids[0]))
['[SEP]', '[CLS]', '北', '京', '是', '中', '国', '的', '首', '都', '[SEP]']
```
**Note: Please use BertTokenizer for the model vocabulary. DO NOT use original BartTokenizer.**
## Citation
```bibtex
@article{shao2021cpt,
title={CPT: A Pre-Trained Unbalanced Transformer for Both Chinese Language Understanding and Generation},
author={Yunfan Shao and Zhichao Geng and Yitao Liu and Junqi Dai and Fei Yang and Li Zhe and Hujun Bao and Xipeng Qiu},
journal={arXiv preprint arXiv:2109.05729},
year={2021}
}
```
|
leo123/BERT-Preguntas-Respuestas-Posgrados | b4e6cabe7b6b1abd9374b01283b24592c749a068 | 2022-07-14T23:20:25.000Z | [
"pytorch",
"bert",
"question-answering",
"transformers",
"license:apache-2.0",
"autotrain_compatible"
] | question-answering | false | leo123 | null | leo123/BERT-Preguntas-Respuestas-Posgrados | 875 | null | transformers | 1,868 | ---
license: apache-2.0
---
|
anas-awadalla/bert-tiny-finetuned-squad | a40bd3413dd077c303653a88626b66b73de0ee04 | 2022-05-21T08:11:40.000Z | [
"pytorch",
"tensorboard",
"bert",
"question-answering",
"dataset:squad",
"transformers",
"generated_from_trainer",
"license:mit",
"model-index",
"autotrain_compatible"
] | question-answering | false | anas-awadalla | null | anas-awadalla/bert-tiny-finetuned-squad | 874 | null | transformers | 1,869 | ---
license: mit
tags:
- generated_from_trainer
datasets:
- squad
model-index:
- name: bert-tiny-finetuned-squad
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. -->
# bert-tiny-finetuned-squad
This model is a fine-tuned version of [prajjwal1/bert-tiny](https://huggingface.co/prajjwal1/bert-tiny) on the squad dataset.
## 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: 3e-05
- train_batch_size: 64
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2.0
### Training results
### Framework versions
- Transformers 4.17.0
- Pytorch 1.11.0+cu113
- Datasets 2.0.0
- Tokenizers 0.11.6
|
shibing624/code-autocomplete-distilgpt2-python | 4a0986fce0baf2b583080207b8539d3fa62002d7 | 2022-02-15T07:18:50.000Z | [
"pytorch",
"gpt2",
"text-generation",
"en",
"transformers",
"code",
"autocomplete",
"license:apache-2.0"
] | text-generation | false | shibing624 | null | shibing624/code-autocomplete-distilgpt2-python | 874 | 7 | transformers | 1,870 | ---
language:
- en
tags:
- code
- autocomplete
- pytorch
- en
license: "apache-2.0"
---
# GPT2 for Code AutoComplete Model
code-autocomplete, a code completion plugin for Python.
**code-autocomplete** can automatically complete the code of lines and blocks with GPT2.
## Usage
Open source repo：[code-autocomplete](https://github.com/shibing624/code-autocomplete)，support GPT2 model, usage：
```python
from autocomplete.gpt2_coder import GPT2Coder
m = GPT2Coder("shibing624/code-autocomplete-distilgpt2-python")
print(m.generate('import torch.nn as')[0])
```
Also, use huggingface/transformers：
*Please use 'GPT2' related functions to load this model!*
```python
import os
from transformers import GPT2Tokenizer, GPT2LMHeadModel
os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
tokenizer = GPT2Tokenizer.from_pretrained("shibing624/code-autocomplete-distilgpt2-python")
model = GPT2LMHeadModel.from_pretrained("shibing624/code-autocomplete-distilgpt2-python")
prompts = [
"""from torch import nn
class LSTM(Module):
def __init__(self, *,
n_tokens: int,
embedding_size: int,
hidden_size: int,
n_layers: int):""",
"""import numpy as np
import torch
import torch.nn as""",
"import java.util.ArrayList",
"def factorial(n):",
]
for prompt in prompts:
input_ids = tokenizer.encode(prompt, add_special_tokens=False, return_tensors='pt')
outputs = model.generate(input_ids=input_ids,
max_length=64 + len(prompt),
temperature=1.0,
top_k=50,
top_p=0.95,
repetition_penalty=1.0,
do_sample=True,
num_return_sequences=1,
length_penalty=2.0,
early_stopping=True)
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(decoded)
print("=" * 20)
```
output:
```shell
from torch import nn
class LSTM(Module):
def __init__(self, *,
n_tokens: int,
embedding_size: int,
hidden_size: int,
n_layers: int):
self.embedding_size = embedding_size
====================
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
```
Model files：
```
code-autocomplete-distilgpt2-python
├── config.json
├── merges.txt
├── pytorch_model.bin
├── special_tokens_map.json
├── tokenizer_config.json
└── vocab.json
```
### Train data
#### pytorch_awesome projects source code
download [code-autocomplete](https://github.com/shibing624/code-autocomplete),
```shell
cd autocomplete
python create_dataset.py
```
If you want train code-autocomplete GPT2 model，refer [https://github.com/shibing624/code-autocomplete/blob/main/autocomplete/gpt2_coder.py](https://github.com/shibing624/code-autocomplete/blob/main/autocomplete/gpt2_coder.py)
### About GPT2
Test the whole generation capabilities here: https://transformer.huggingface.co/doc/gpt2-large
Pretrained model on English language using a causal language modeling (CLM) objective. It was introduced in
[this paper](https://d4mucfpksywv.cloudfront.net/better-language-models/language_models_are_unsupervised_multitask_learners.pdf)
and first released at [this page](https://openai.com/blog/better-language-models/).
Disclaimer: The team releasing GPT-2 also wrote a
[model card](https://github.com/openai/gpt-2/blob/master/model_card.md) for their model. Content from this model card
has been written by the Hugging Face team to complete the information they provided and give specific examples of bias.
## Citation
```latex
@misc{code-autocomplete,
author = {Xu Ming},
title = {code-autocomplete: Code AutoComplete with GPT model},
year = {2022},
publisher = {GitHub},
journal = {GitHub repository},
url = {https://github.com/shibing624/code-autocomplete},
}
```
|
squeezebert/squeezebert-mnli-headless | d90683f6cb548ca6019ce3366d03f8652d836b7e | 2020-12-11T22:02:10.000Z | [
"pytorch",
"squeezebert",
"arxiv:2006.11316",
"arxiv:1904.00962",
"transformers"
] | null | false | squeezebert | null | squeezebert/squeezebert-mnli-headless | 874 | null | transformers | 1,871 | language: en
license: bsd
datasets:
- bookcorpus
- wikipedia
---
# SqueezeBERT pretrained model
This model, `squeezebert-mnli-headless`, has been pretrained for the English language using a masked language modeling (MLM) and Sentence Order Prediction (SOP) objective and finetuned on the [Multi-Genre Natural Language Inference (MNLI)](https://cims.nyu.edu/~sbowman/multinli/) dataset. This is a "headless" model with the final classification layer removed, and this will allow Transformers to automatically reinitialize the final classification layer before you begin finetuning on your data.
SqueezeBERT was introduced in [this paper](https://arxiv.org/abs/2006.11316). This model is case-insensitive. The model architecture is similar to BERT-base, but with the pointwise fully-connected layers replaced with [grouped convolutions](https://blog.yani.io/filter-group-tutorial/).
The authors found that SqueezeBERT is 4.3x faster than `bert-base-uncased` on a Google Pixel 3 smartphone.
## Pretraining
### Pretraining data
- [BookCorpus](https://yknzhu.wixsite.com/mbweb), a dataset consisting of thousands of unpublished books
- [English Wikipedia](https://en.wikipedia.org/wiki/English_Wikipedia)
### Pretraining procedure
The model is pretrained using the Masked Language Model (MLM) and Sentence Order Prediction (SOP) tasks.
(Author's note: If you decide to pretrain your own model, and you prefer to train with MLM only, that should work too.)
From the SqueezeBERT paper:
> We pretrain SqueezeBERT from scratch (without distillation) using the [LAMB](https://arxiv.org/abs/1904.00962) optimizer, and we employ the hyperparameters recommended by the LAMB authors: a global batch size of 8192, a learning rate of 2.5e-3, and a warmup proportion of 0.28. Following the LAMB paper's recommendations, we pretrain for 56k steps with a maximum sequence length of 128 and then for 6k steps with a maximum sequence length of 512.
## Finetuning
The SqueezeBERT paper presents 2 approaches to finetuning the model:
- "finetuning without bells and whistles" -- after pretraining the SqueezeBERT model, finetune it on each GLUE task
- "finetuning with bells and whistles" -- after pretraining the SqueezeBERT model, finetune it on a MNLI with distillation from a teacher model. Then, use the MNLI-finetuned SqueezeBERT model as a student model to finetune on each of the other GLUE tasks (e.g. RTE, MRPC, …) with distillation from a task-specific teacher model.
A detailed discussion of the hyperparameters used for finetuning is provided in the appendix of the [SqueezeBERT paper](https://arxiv.org/abs/2006.11316).
Note that finetuning SqueezeBERT with distillation is not yet implemented in this repo. If the author (Forrest Iandola - [email protected]) gets enough encouragement from the user community, he will add example code to Transformers for finetuning SqueezeBERT with distillation.
This model, `squeezebert/squeezebert-mnli-headless`, is the "finetuned with bells and whistles" MNLI-finetuned SqueezeBERT model. In this particular model, we have removed the final classification layer -- in other words, it is "headless." We recommend using this model if you intend to finetune the model on your own data. Using this model means that your final layer will automatically be reinitialized when you start finetuning on your data.
### How to finetune
To try finetuning SqueezeBERT on the [MRPC](https://www.microsoft.com/en-us/download/details.aspx?id=52398) text classification task, you can run the following command:
```
./utils/download_glue_data.py
python examples/text-classification/run_glue.py \
--model_name_or_path squeezebert-base-headless \
--task_name mrpc \
--data_dir ./glue_data/MRPC \
--output_dir ./models/squeezebert_mrpc \
--overwrite_output_dir \
--do_train \
--do_eval \
--num_train_epochs 10 \
--learning_rate 3e-05 \
--per_device_train_batch_size 16 \
--save_steps 20000
```
## BibTeX entry and citation info
```
@article{2020_SqueezeBERT,
author = {Forrest N. Iandola and Albert E. Shaw and Ravi Krishna and Kurt W. Keutzer},
title = {{SqueezeBERT}: What can computer vision teach NLP about efficient neural networks?},
journal = {arXiv:2006.11316},
year = {2020}
}
```
|
akhooli/gpt2-small-arabic-poetry | 7c384f4f774a83aae4c66251050528c4b33b36d3 | 2021-08-07T08:06:39.000Z | [
"pytorch",
"jax",
"gpt2",
"text-generation",
"ar",
"dataset:Arabic poetry from several eras",
"transformers"
] | text-generation | false | akhooli | null | akhooli/gpt2-small-arabic-poetry | 873 | 3 | transformers | 1,872 | ---
language: "ar"
tags:
- text-generation
datasets:
- Arabic poetry from several eras
---
# GPT2-Small-Arabic-Poetry
## Model description
Fine-tuned model of Arabic poetry dataset based on gpt2-small-arabic.
## Intended uses & limitations
#### How to use
An example is provided in this [colab notebook](https://colab.research.google.com/drive/1mRl7c-5v-Klx27EEAEOAbrfkustL4g7a?usp=sharing).
#### Limitations and bias
Both the GPT2-small-arabic (trained on Arabic Wikipedia) and this model have several limitations in terms of coverage and training performance.
Use them as demonstrations or proof of concepts but not as production code.
## Training data
This pretrained model used the [Arabic Poetry dataset](https://www.kaggle.com/ahmedabelal/arabic-poetry) from 9 different eras with a total of around 40k poems.
The dataset was trained (fine-tuned) based on the [gpt2-small-arabic](https://huggingface.co/akhooli/gpt2-small-arabic) transformer model.
## Training procedure
Training was done using [Simple Transformers](https://github.com/ThilinaRajapakse/simpletransformers) library on Kaggle, using free GPU.
## Eval results
Final perplexity reached ws 76.3, loss: 4.33
### BibTeX entry and citation info
```bibtex
@inproceedings{Abed Khooli,
year={2020}
}
```
|
nboost/pt-bert-large-msmarco | 654bca99ecec6faf688b274cb9c99333e9251c3f | 2021-05-20T01:25:29.000Z | [
"pytorch",
"jax",
"onnx",
"bert",
"transformers"
] | null | false | nboost | null | nboost/pt-bert-large-msmarco | 873 | 1 | transformers | 1,873 | Entry not found |
MaryaAI/opus-mt-ar-en-finetuned-ar-to-en | bc61a529dce1ad153ed501e41997e97d44b24262 | 2021-09-07T07:26:24.000Z | [
"pytorch",
"tensorboard",
"marian",
"text2text-generation",
"dataset:opus_wikipedia",
"transformers",
"generated_from_trainer",
"model-index",
"autotrain_compatible"
] | text2text-generation | false | MaryaAI | null | MaryaAI/opus-mt-ar-en-finetuned-ar-to-en | 872 | null | transformers | 1,874 | ---
tags:
- generated_from_trainer
datasets:
- opus_wikipedia
model-index:
- name: opus-mt-ar-en-finetuned-ar-to-en
results:
- task:
name: Sequence-to-sequence Language Modeling
type: text2text-generation
dataset:
name: opus_wikipedia
type: opus_wikipedia
args: ar-en
---
<!-- 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. -->
# opus-mt-ar-en-finetuned-ar-to-en
This model is a fine-tuned version of [Helsinki-NLP/opus-mt-ar-en](https://huggingface.co/Helsinki-NLP/opus-mt-ar-en) on the opus_wikipedia dataset.
## 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: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1
- mixed_precision_training: Native AMP
### Framework versions
- Transformers 4.10.0
- Pytorch 1.9.0+cu102
- Datasets 1.11.0
- Tokenizers 0.10.3
|
microsoft/tapex-base | 968109c940c8b270a3eaec1532d596ba6c923b6a | 2022-05-17T08:25:49.000Z | [
"pytorch",
"bart",
"text2text-generation",
"en",
"arxiv:2107.07653",
"transformers",
"tapex",
"table-question-answering",
"license:mit",
"autotrain_compatible"
] | table-question-answering | false | microsoft | null | microsoft/tapex-base | 871 | 4 | transformers | 1,875 | ---
language: en
tags:
- tapex
- table-question-answering
license: mit
---
# TAPEX (base-sized model)
TAPEX was proposed in [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou. The original repo can be found [here](https://github.com/microsoft/Table-Pretraining).
## Model description
TAPEX (**Ta**ble **P**re-training via **Ex**ecution) is a conceptually simple and empirically powerful pre-training approach to empower existing models with *table reasoning* skills. TAPEX realizes table pre-training by learning a neural SQL executor over a synthetic corpus, which is obtained by automatically synthesizing executable SQL queries.
TAPEX is based on the BART architecture, the transformer encoder-encoder (seq2seq) model with a bidirectional (BERT-like) encoder and an autoregressive (GPT-like) decoder.
## Intended Uses
You can use the raw model for simulating neural SQL execution, i.e., employ TAPEX to execute a SQL query on a given table. However, the model is mostly meant to be fine-tuned on a supervised dataset. Currently TAPEX can be fine-tuned to tackle table question answering tasks and table fact verification tasks. See the [model hub](https://huggingface.co/models?search=tapex) to look for fine-tuned versions on a task that interests you.
### How to Use
Here is how to use this model in transformers:
```python
from transformers import TapexTokenizer, BartForConditionalGeneration
import pandas as pd
tokenizer = TapexTokenizer.from_pretrained("microsoft/tapex-base")
model = BartForConditionalGeneration.from_pretrained("microsoft/tapex-base")
data = {
"year": [1896, 1900, 1904, 2004, 2008, 2012],
"city": ["athens", "paris", "st. louis", "athens", "beijing", "london"]
}
table = pd.DataFrame.from_dict(data)
# tapex accepts uncased input since it is pre-trained on the uncased corpus
query = "select year where city = beijing"
encoding = tokenizer(table=table, query=query, return_tensors="pt")
outputs = model.generate(**encoding)
print(tokenizer.batch_decode(outputs, skip_special_tokens=True))
# ['2008']
```
### How to Fine-tuning
Please find the fine-tuning script [here](https://github.com/SivilTaram/transformers/tree/add_tapex_bis/examples/research_projects/tapex).
### BibTeX entry and citation info
```bibtex
@inproceedings{
liu2022tapex,
title={{TAPEX}: Table Pre-training via Learning a Neural {SQL} Executor},
author={Qian Liu and Bei Chen and Jiaqi Guo and Morteza Ziyadi and Zeqi Lin and Weizhu Chen and Jian-Guang Lou},
booktitle={International Conference on Learning Representations},
year={2022},
url={https://openreview.net/forum?id=O50443AsCP}
}
``` |
google/realm-cc-news-pretrained-scorer | a009929c7c945e823f1e0c4ee0ea3c737606a6de | 2022-01-06T06:23:03.000Z | [
"pytorch",
"realm",
"en",
"transformers",
"license:apache-2.0"
] | null | false | google | null | google/realm-cc-news-pretrained-scorer | 868 | null | transformers | 1,876 | ---
language: en
license: apache-2.0
---
# realm-cc-news-pretrained-scorer
## Model description
The REALM checkpoint pretrained with CC-News as target corpus and Wikipedia as knowledge corpus, converted from the TF checkpoint provided by Google Language.
The original paper, code, and checkpoints can be found [here](https://github.com/google-research/language/tree/master/language/realm).
## Usage
```python
from transformers import RealmScorer
scorer = RealmScorer.from_pretrained("qqaatw/realm-cc-news-pretrained-scorer")
```
|
stefan-it/bort | 3afaf7981024b80cfa229b2e271323f8c0aa1c6b | 2021-05-20T07:14:56.000Z | [
"pytorch",
"tf",
"jax",
"bert",
"fill-mask",
"transformers",
"autotrain_compatible"
] | fill-mask | false | stefan-it | null | stefan-it/bort | 863 | null | transformers | 1,877 | Entry not found |
lgrobol/flaubert-minuscule | b2761368313c6c178ae6d3ac10332632e9af8170 | 2021-08-17T13:19:07.000Z | [
"pytorch",
"flaubert",
"fill-mask",
"transformers",
"autotrain_compatible"
] | fill-mask | false | lgrobol | null | lgrobol/flaubert-minuscule | 861 | null | transformers | 1,878 | FlauBERT-minuscule
==================
A ridiculously small model for testing purposes. |
microsoft/resnet-101 | c2bf50f68263a35f102eb5c84ca91fc7352ceff3 | 2022-07-01T17:33:19.000Z | [
"pytorch",
"tf",
"resnet",
"image-classification",
"dataset:imagenet-1k",
"arxiv:1512.03385",
"transformers",
"vision",
"license:apache-2.0"
] | image-classification | false | microsoft | null | microsoft/resnet-101 | 860 | 2 | transformers | 1,879 | ---
license: apache-2.0
tags:
- vision
- image-classification
datasets:
- imagenet-1k
---
# ResNet-101 v1.5
ResNet model pre-trained on ImageNet-1k at resolution 224x224. It was introduced in the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by He et al.
Disclaimer: The team releasing ResNet did not write a model card for this model so this model card has been written by the Hugging Face team.
## Model description
ResNet (Residual Network) is a convolutional neural network that democratized the concepts of residual learning and skip connections. This enables to train much deeper models.
This is ResNet v1.5, which differs from the original model: in the bottleneck blocks which require downsampling, v1 has stride = 2 in the first 1x1 convolution, whereas v1.5 has stride = 2 in the 3x3 convolution. This difference makes ResNet50 v1.5 slightly more accurate (\~0.5% top1) than v1, but comes with a small performance drawback (~5% imgs/sec) according to [Nvidia](https://catalog.ngc.nvidia.com/orgs/nvidia/resources/resnet_50_v1_5_for_pytorch).
## Intended uses & limitations
You can use the raw model for image classification. See the [model hub](https://huggingface.co/models?search=resnet) to look for
fine-tuned versions on a task that interests you.
### How to use
Here is how to use this model to classify an image of the COCO 2017 dataset into one of the 1,000 ImageNet classes:
```python
from transformers import AutoFeatureExtractor, ResNetForImageClassification
import torch
from datasets import load_dataset
dataset = load_dataset("huggingface/cats-image")
image = dataset["test"]["image"][0]
feature_extractor = AutoFeatureExtractor.from_pretrained("microsoft/resnet-101")
model = ResNetForImageClassification.from_pretrained("microsoft/resnet-101")
inputs = feature_extractor(image, return_tensors="pt")
with torch.no_grad():
logits = model(**inputs).logits
# model predicts one of the 1000 ImageNet classes
predicted_label = logits.argmax(-1).item()
print(model.config.id2label[predicted_label])
```
For more code examples, we refer to the [documentation](https://huggingface.co/docs/transformers/main/en/model_doc/resnet).
### BibTeX entry and citation info
```bibtex
@inproceedings{he2016deep,
title={Deep residual learning for image recognition},
author={He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian},
booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
pages={770--778},
year={2016}
}
```
|
SIC98/GPT2-python-code-generator | 525aec7829d6b9606e01f979d58cf4125fb906e5 | 2021-05-21T11:13:58.000Z | [
"pytorch",
"jax",
"gpt2",
"text-generation",
"transformers"
] | text-generation | false | SIC98 | null | SIC98/GPT2-python-code-generator | 859 | 3 | transformers | 1,880 | Github
- https://github.com/SIC98/GPT2-python-code-generator |
Helsinki-NLP/opus-mt-is-en | 2334bf160857f518815cd97a0b7a3c5e81b7fa2e | 2021-09-09T22:12:09.000Z | [
"pytorch",
"marian",
"text2text-generation",
"is",
"en",
"transformers",
"translation",
"license:apache-2.0",
"autotrain_compatible"
] | translation | false | Helsinki-NLP | null | Helsinki-NLP/opus-mt-is-en | 858 | 1 | transformers | 1,881 | ---
tags:
- translation
license: apache-2.0
---
### opus-mt-is-en
* source languages: is
* target languages: en
* OPUS readme: [is-en](https://github.com/Helsinki-NLP/OPUS-MT-train/blob/master/models/is-en/README.md)
* dataset: opus
* model: transformer-align
* pre-processing: normalization + SentencePiece
* download original weights: [opus-2019-12-18.zip](https://object.pouta.csc.fi/OPUS-MT-models/is-en/opus-2019-12-18.zip)
* test set translations: [opus-2019-12-18.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/is-en/opus-2019-12-18.test.txt)
* test set scores: [opus-2019-12-18.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/is-en/opus-2019-12-18.eval.txt)
## Benchmarks
| testset | BLEU | chr-F |
|-----------------------|-------|-------|
| Tatoeba.is.en | 51.4 | 0.672 |
|
Lalita/marianmt-th-zh_cn | bab48cedd044cf1c5b9064943bf52a78728c9721 | 2021-06-29T14:06:47.000Z | [
"pytorch",
"marian",
"text2text-generation",
"transformers",
"translation",
"torch==1.8.0",
"autotrain_compatible"
] | translation | false | Lalita | null | Lalita/marianmt-th-zh_cn | 858 | null | transformers | 1,882 | ---
tags:
- translation
- torch==1.8.0
widget:
- text: "Inference Unavailable"
---
### marianmt-th-zh_cn
* source languages: th
* target languages: zh_cn
* dataset:
* model: transformer-align
* pre-processing: normalization + SentencePiece
* test set scores: 15.53
## Training
Training scripts from [LalitaDeelert/NLP-ZH_TH-Project](https://github.com/LalitaDeelert/NLP-ZH_TH-Project). Experiments tracked at [cstorm125/marianmt-th-zh_cn](https://wandb.ai/cstorm125/marianmt-th-zh_cn).
```
export WANDB_PROJECT=marianmt-th-zh_cn
python train_model.py --input_fname ../data/v1/Train.csv \\\\\\\\
\\\\t--output_dir ../models/marianmt-th-zh_cn \\\\\\\\
\\\\t--source_lang th --target_lang zh \\\\\\\\
\\\\t--metric_tokenize zh --fp16
```
## Usage
```
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("Lalita/marianmt-zh_cn-th")
model = AutoModelForSeq2SeqLM.from_pretrained("Lalita/marianmt-zh_cn-th").cpu()
src_text = [
'ฉันรักคุณ',
'ฉันอยากกินข้าว',
]
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
print([tokenizer.decode(t, skip_special_tokens=True) for t in translated])
> ['我爱你', '我想吃饭。']
```
## Requirements
```
transformers==4.6.0
torch==1.8.0
``` |
savasy/bert-base-turkish-squad | 0309abfbf39abc803db200667d25a003affd5112 | 2021-05-20T04:56:01.000Z | [
"pytorch",
"jax",
"bert",
"question-answering",
"tr",
"transformers",
"autotrain_compatible"
] | question-answering | false | savasy | null | savasy/bert-base-turkish-squad | 857 | 5 | transformers | 1,883 | ---
language: tr
---
# Turkish SQuAD Model : Question Answering
I fine-tuned Turkish-Bert-Model for Question-Answering problem with Turkish version of SQuAD; TQuAD
* BERT-base: https://huggingface.co/dbmdz/bert-base-turkish-uncased
* TQuAD dataset: https://github.com/TQuad/turkish-nlp-qa-dataset
# Training Code
```
!python3 run_squad.py \
--model_type bert \
--model_name_or_path dbmdz/bert-base-turkish-uncased\
--do_train \
--do_eval \
--train_file trainQ.json \
--predict_file dev1.json \
--per_gpu_train_batch_size 12 \
--learning_rate 3e-5 \
--num_train_epochs 5.0 \
--max_seq_length 384 \
--doc_stride 128 \
--output_dir "./model"
```
# Example Usage
> Load Model
```
from transformers import AutoTokenizer, AutoModelForQuestionAnswering, pipeline
import torch
tokenizer = AutoTokenizer.from_pretrained("savasy/bert-base-turkish-squad")
model = AutoModelForQuestionAnswering.from_pretrained("savasy/bert-base-turkish-squad")
nlp=pipeline("question-answering", model=model, tokenizer=tokenizer)
```
> Apply the model
```
sait="ABASIYANIK, Sait Faik. Hikayeci (Adapazarı 23 Kasım 1906-İstanbul 11 Mayıs 1954). \
İlk öğrenimine Adapazarı’nda Rehber-i Terakki Mektebi’nde başladı. İki yıl kadar Adapazarı İdadisi’nde okudu.\
İstanbul Erkek Lisesi’nde devam ettiği orta öğrenimini Bursa Lisesi’nde tamamladı (1928). İstanbul Edebiyat \
Fakültesi’ne iki yıl devam ettikten sonra babasının isteği üzerine iktisat öğrenimi için İsviçre’ye gitti. \
Kısa süre sonra iktisat öğrenimini bırakarak Lozan’dan Grenoble’a geçti. Üç yıl başıboş bir edebiyat öğrenimi \
gördükten sonra babası tarafından geri çağrıldı (1933). Bir müddet Halıcıoğlu Ermeni Yetim Mektebi'nde Türkçe \
gurup dersleri öğretmenliği yaptı. Ticarete atıldıysa da tutunamadı. Bir ay Haber gazetesinde adliye muhabirliği\
yaptı (1942). Babasının ölümü üzerine aileden kalan emlakin geliri ile avare bir hayata başladı. Evlenemedi.\
Yazları Burgaz adasındaki köşklerinde, kışları Şişli’deki apartmanlarında annesi ile beraber geçen bu fazla \
içkili bohem hayatı ömrünün sonuna kadar sürdü."
print(nlp(question="Ne zaman avare bir hayata başladı?", context=sait))
print(nlp(question="Sait Faik hangi Lisede orta öğrenimini tamamladı?", context=sait))
```
```
# Ask your self ! type your question
print(nlp(question="...?", context=sait))
```
Check My other Model
https://huggingface.co/savasy
|
sshleifer/distilbart-xsum-12-3 | 1d2bfbc16dcdd28720f9f1d37be764e5cc5c78c8 | 2021-06-14T07:57:16.000Z | [
"pytorch",
"jax",
"bart",
"text2text-generation",
"en",
"dataset:cnn_dailymail",
"dataset:xsum",
"transformers",
"summarization",
"license:apache-2.0",
"autotrain_compatible"
] | summarization | false | sshleifer | null | sshleifer/distilbart-xsum-12-3 | 856 | 1 | transformers | 1,884 | ---
language: en
tags:
- summarization
license: apache-2.0
datasets:
- cnn_dailymail
- xsum
thumbnail: https://huggingface.co/front/thumbnails/distilbart_medium.png
---
### Usage
This checkpoint should be loaded into `BartForConditionalGeneration.from_pretrained`. See the [BART docs](https://huggingface.co/transformers/model_doc/bart.html?#transformers.BartForConditionalGeneration) for more information.
### Metrics for DistilBART models
| Model Name | MM Params | Inference Time (MS) | Speedup | Rouge 2 | Rouge-L |
|:---------------------------|------------:|----------------------:|----------:|----------:|----------:|
| distilbart-xsum-12-1 | 222 | 90 | 2.54 | 18.31 | 33.37 |
| distilbart-xsum-6-6 | 230 | 132 | 1.73 | 20.92 | 35.73 |
| distilbart-xsum-12-3 | 255 | 106 | 2.16 | 21.37 | 36.39 |
| distilbart-xsum-9-6 | 268 | 136 | 1.68 | 21.72 | 36.61 |
| bart-large-xsum (baseline) | 406 | 229 | 1 | 21.85 | 36.50 |
| distilbart-xsum-12-6 | 306 | 137 | 1.68 | 22.12 | 36.99 |
| bart-large-cnn (baseline) | 406 | 381 | 1 | 21.06 | 30.63 |
| distilbart-12-3-cnn | 255 | 214 | 1.78 | 20.57 | 30.00 |
| distilbart-12-6-cnn | 306 | 307 | 1.24 | 21.26 | 30.59 |
| distilbart-6-6-cnn | 230 | 182 | 2.09 | 20.17 | 29.70 |
|
bhadresh-savani/distilbert-base-uncased-sentiment-sst2 | b91676624bfec8bb96d31c4d0f1b13a491ebe65c | 2022-06-15T11:48:33.000Z | [
"pytorch",
"tf",
"jax",
"distilbert",
"text-classification",
"en",
"dataset:sst2",
"transformers",
"license:apache-2.0"
] | text-classification | false | bhadresh-savani | null | bhadresh-savani/distilbert-base-uncased-sentiment-sst2 | 854 | null | transformers | 1,885 | ---
language: en
license: apache-2.0
datasets:
- sst2
---
# distilbert-base-uncased-sentiment-sst2
This model will be able to identify positivity or negativity present in the sentence
## Dataset:
The Stanford Sentiment Treebank from GLUE
## Results:
```
***** eval metrics *****
epoch = 3.0
eval_accuracy = 0.9094
eval_loss = 0.3514
eval_runtime = 0:00:03.60
eval_samples = 872
eval_samples_per_second = 242.129
eval_steps_per_second = 30.266
``` |
tanlq/vit-base-patch16-224-in21k-finetuned-cifar10 | b180bcaf51fdf309391ae08a72494bf9fbf7d64a | 2022-04-04T08:20:16.000Z | [
"pytorch",
"vit",
"image-classification",
"dataset:cifar10",
"transformers",
"generated_from_trainer",
"license:apache-2.0",
"model-index"
] | image-classification | false | tanlq | null | tanlq/vit-base-patch16-224-in21k-finetuned-cifar10 | 854 | null | transformers | 1,886 | ---
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- cifar10
metrics:
- accuracy
model-index:
- name: vit-base-patch16-224-in21k-finetuned-cifar10
results:
- task:
name: Image Classification
type: image-classification
dataset:
name: cifar10
type: cifar10
args: plain_text
metrics:
- name: Accuracy
type: accuracy
value: 0.9875
---
<!-- 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. -->
# vit-base-patch16-224-in21k-finetuned-cifar10
This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the cifar10 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0503
- Accuracy: 0.9875
## 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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3
### Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 0.3118 | 1.0 | 1562 | 0.1135 | 0.9778 |
| 0.2717 | 2.0 | 3124 | 0.0619 | 0.9867 |
| 0.1964 | 3.0 | 4686 | 0.0503 | 0.9875 |
### Framework versions
- Transformers 4.18.0.dev0
- Pytorch 1.11.0
- Datasets 2.0.0
- Tokenizers 0.11.6
|
gorkemgoknar/gpt2-small-turkish | cecfcbf3dfbb3c9df280386790c0ac45d21ad9d9 | 2021-09-22T08:29:21.000Z | [
"pytorch",
"jax",
"gpt2",
"text-generation",
"tr",
"dataset:wikipedia-turkish",
"transformers",
"turkish",
"license:apache-2.0"
] | text-generation | false | gorkemgoknar | null | gorkemgoknar/gpt2-small-turkish | 852 | 1 | transformers | 1,887 | ---
language:
- tr
thumbnail:
tags:
- gpt2
- turkish
license: apache-2.0
datasets:
- wikipedia-turkish
metrics:
- perplexity
- accuracy
widget:
- text: Bu yazıyı bir bilgisayar yazdı. Yazarken
context: ''
- text: İnternete kolay erişim sayesinde dünya daha da küçüldü. Bunun sonucunda
context: ''
---
# Turkish GPT2 Model Finetuned
# Türkçe GPT2 Modeli
## Model description
This is a GPT2-Small English based model finetuned and additionaly trainied with Wikipedia Articles in Turkish as of 28-10-2020
Live demo based on this work at : https://www.metayazar.com/
Fine tuned writer on this model: https://huggingface.co/gorkemgoknar/gpt2-turkish-writer
Work has been done on Pierre Guillou tutorial as on this page.
(https://github.com/piegu/fastai-projects/blob/master/finetuning-English-GPT2-any-language-Portuguese-HuggingFace-fastaiv2.ipynb)
Code is converted to work with Fastai 2.X .
Using Google Colab for training.
Additional tutorial and source will be in https://github.com/gorkemgoknar in later stage.
Current accuracy 33 % , Perplexity : 51.88
Models are available:
* [gpt2-small-tuned-tr] (https://huggingface.co/gorkemgoknar/gpt2-small-turkish)
* [gpt2-small-turkish-writer] (https://huggingface.co/gorkemgoknar/gpt2-turkish-writer)
## Intended uses & limitations
#### How to use
#### Install
```python
from transformers import AutoTokenizer, AutoModelWithLMHead
import torch
tokenizer = AutoTokenizer.from_pretrained("gorkemgoknar/gpt2-small-turkish")
model = AutoModelWithLMHead.from_pretrained("gorkemgoknar/gpt2-small-turkish")
# Get sequence length max of 1024
tokenizer.model_max_length=1024
model.eval() # disable dropout (or leave in train mode to finetune)
```
#### Generate 1 word
```python
# input sequence
text = "Bu yazıyı bilgisayar yazdı."
inputs = tokenizer(text, return_tensors="pt")
# model output
outputs = model(**inputs, labels=inputs["input_ids"])
loss, logits = outputs[:2]
predicted_index = torch.argmax(logits[0, -1, :]).item()
predicted_text = tokenizer.decode([predicted_index])
# results
print('input text:', text)
print('predicted text:', predicted_text)
# input text:
# predicted text:
```
#### Generate Full Sequence
```python
# input sequence
text = "Bu yazıyı bilgisayar yazdı."
inputs = tokenizer(text, return_tensors="pt")
# model output using Top-k sampling text generation method
sample_outputs = model.generate(inputs.input_ids,
pad_token_id=50256,
do_sample=True,
max_length=50, # put the token number you want
top_k=40,
num_return_sequences=1)
# generated sequence
for i, sample_output in enumerate(sample_outputs):
print(">> Generated text {}\\\\
\\\\
{}".format(i+1, tokenizer.decode(sample_output.tolist())))
# >> Generated text
#
```
#### Limitations and bias
The training data used for this model come from Turkish Wikipedia. We know it contains a lot of unfiltered content from the internet, which is far from neutral.
## Training data
Wikipedia Turkish article dump as of 28-10-2020
## Training procedure
## Eval results
| epoch\\\\t|train_loss\\\\t|valid_loss\\\\t|accuracy\\\\t|perplexity\\\\t|time |
| ----- | -------- |--------- | ---------- | --------- | ----- |
|0\\\\t|4.777015\\\\t|4.621834\\\\t|0.292547\\\\t|101.680367\\\\t|2:42:05|
|1\\\\t|4.509412\\\\t|4.403999\\\\t|0.305574\\\\t|81.777267\\\\t|1:09:38|
|2\\\\t|4.169529\\\\t|4.120755\\\\t|0.324908\\\\t|61.605747\\\\t|1:07:45|
|3\\\\t|4.293973\\\\t|4.177899\\\\t|0.317211\\\\t|65.228653\\\\t|1:07:02|
|4\\\\t|4.049848\\\\t|3.949103\\\\t|0.338347\\\\t|51.888783\\\\t|1:05:53|
#Epoch 0 on Tesla T4, others on V100
```
|
uclanlp/visualbert-nlvr2 | 2cb80570d2326bbb1f3a954f967a1cd5bed949b6 | 2021-05-31T11:09:59.000Z | [
"pytorch",
"visual_bert",
"transformers"
] | null | false | uclanlp | null | uclanlp/visualbert-nlvr2 | 852 | null | transformers | 1,888 | Entry not found |
KBLab/sentence-bert-swedish-cased | e5e754ac75b8dddc1c15f52e11ef7d326792fd1e | 2022-07-28T14:18:47.000Z | [
"pytorch",
"bert",
"feature-extraction",
"arxiv:2004.09813",
"sentence-transformers",
"sentence-similarity",
"transformers"
] | sentence-similarity | false | KBLab | null | KBLab/sentence-bert-swedish-cased | 851 | 3 | sentence-transformers | 1,889 | ---
pipeline_tag: sentence-similarity
tags:
- sentence-transformers
- feature-extraction
- sentence-similarity
- transformers
widget:
- source_sentence: "Mannen åt mat."
sentences:
- "Han förtärde en närande och nyttig måltid."
- "Det var ett sunkigt hak med ganska gott käk."
- "Han inmundigade middagen tillsammans med ett glas rödvin."
- "Potatischips är jättegoda."
- "Tryck på knappen för att få tala med kundsupporten."
example_title: "Mat"
- source_sentence: "Kan jag deklarera digitalt från utlandet?"
sentences:
- "Du som befinner dig i utlandet kan deklarera digitalt på flera olika sätt."
- "Du som har kvarskatt att betala ska göra en inbetalning till ditt skattekonto."
- "Efter att du har deklarerat går vi igenom uppgifterna i din deklaration och räknar ut din skatt."
- "I din deklaration som du får från oss har vi räknat ut vad du ska betala eller få tillbaka."
- "Tryck på knappen för att få tala med kundsupporten."
example_title: "Skatteverket FAQ"
- source_sentence: "Hon kunde göra bakåtvolter."
sentences:
- "Hon var atletisk."
- "Hon var bra på gymnastik."
- "Hon var inte atletisk."
- "Hon var oförmögen att flippa baklänges."
example_title: "Gymnastik"
---
# KBLab/sentence-bert-swedish-cased
This is a [sentence-transformers](https://www.SBERT.net) model: It maps Swedish sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. This model is a bilingual Swedish-English model trained according to instructions in the paper [Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation](https://arxiv.org/pdf/2004.09813.pdf) and the [documentation](https://www.sbert.net/examples/training/multilingual/README.html) accompanying its companion python package. We have used the strongest available pretrained English Bi-Encoder ([paraphrase-mpnet-base-v2](https://www.sbert.net/docs/pretrained_models.html#sentence-embedding-models)) as a teacher model, and the pretrained Swedish [KB-BERT](https://huggingface.co/KB/bert-base-swedish-cased) as the student model.
A more detailed description of the model can be found in an article we published on the [KBLab blog](https://kb-labb.github.io/posts/2021-08-23-a-swedish-sentence-transformer/).
<!--- Describe your model here -->
## Usage (Sentence-Transformers)
Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
```
pip install -U sentence-transformers
```
Then you can use the model like this:
```python
from sentence_transformers import SentenceTransformer
sentences = ["Det här är en exempelmening", "Varje exempel blir konverterad"]
model = SentenceTransformer('KBLab/sentence-bert-swedish-cased')
embeddings = model.encode(sentences)
print(embeddings)
```
## Usage (HuggingFace Transformers)
Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.
```python
from transformers import AutoTokenizer, AutoModel
import torch
#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
token_embeddings = model_output[0] #First element of model_output contains all token embeddings
input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
# Sentences we want sentence embeddings for
sentences = ['Det här är en exempelmening', 'Varje exempel blir konverterad']
# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('KBLab/sentence-bert-swedish-cased')
model = AutoModel.from_pretrained('KBLab/sentence-bert-swedish-cased')
# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
# Compute token embeddings
with torch.no_grad():
model_output = model(**encoded_input)
# Perform pooling. In this case, max pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
print("Sentence embeddings:")
print(sentence_embeddings)
```
## Evaluation Results
<!--- Describe how your model was evaluated -->
The model was primarily evaluated on [SweParaphrase v1.0](https://spraakbanken.gu.se/en/resources/sweparaphrase). This test set is part of [SuperLim](https://spraakbanken.gu.se/en/resources/superlim) -- a Swedish evaluation suite for natural langage understanding tasks. We calculated Pearson and Spearman correlation between predicted model similarity scores and the human similarity score labels. The model achieved a Pearson correlation coefficient of **0.918** and a Spearman's rank correlation coefficient of **0.911**.
The following code snippet can be used to reproduce the above results:
```python
from sentence_transformers import SentenceTransformer
import pandas as pd
df = pd.read_csv(
"sweparaphrase-dev-165.csv",
sep="\t",
header=None,
names=[
"original_id",
"source",
"type",
"sentence_swe1",
"sentence_swe2",
"score",
"sentence1",
"sentence2",
],
)
model = SentenceTransformer("KBLab/sentence-bert-swedish-cased")
sentences1 = df["sentence_swe1"].tolist()
sentences2 = df["sentence_swe2"].tolist()
# Compute embedding for both lists
embeddings1 = model.encode(sentences1, convert_to_tensor=True)
embeddings2 = model.encode(sentences2, convert_to_tensor=True)
# Compute cosine similarity after normalizing
embeddings1 /= embeddings1.norm(dim=-1, keepdim=True)
embeddings2 /= embeddings2.norm(dim=-1, keepdim=True)
cosine_scores = embeddings1 @ embeddings2.t()
sentence_pair_scores = cosine_scores.diag()
df["model_score"] = sentence_pair_scores.cpu().tolist()
print(df[["score", "model_score"]].corr(method="spearman"))
print(df[["score", "model_score"]].corr(method="pearson"))
```
Examples how to evaluate the model on other test sets of the SuperLim suites can be found on the following links: [evaluate_faq.py](https://github.com/kb-labb/swedish-sbert/blob/main/evaluate_faq.py) (Swedish FAQ), [evaluate_swesat.py](https://github.com/kb-labb/swedish-sbert/blob/main/evaluate_swesat.py) (SweSAT synonyms), [evaluate_supersim.py](https://github.com/kb-labb/swedish-sbert/blob/main/evaluate_supersim.py) (SuperSim).
## Training
An article with more details on data and the model can be found on the [KBLab blog](https://kb-labb.github.io/posts/2021-08-23-a-swedish-sentence-transformer/).
Around 14.6 million sentences from English-Swedish parallel corpuses were used to train the model. Data was sourced from the [Open Parallel Corpus](https://opus.nlpl.eu/) (OPUS) and downloaded via the python package [opustools](https://pypi.org/project/opustools/). Datasets used were: JW300, Europarl, EUbookshop, EMEA, TED2020, Tatoeba and OpenSubtitles.
The model was trained with the parameters:
**DataLoader**:
`torch.utils.data.dataloader.DataLoader` of length 227832 with parameters:
```
{'batch_size': 64, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
```
**Loss**:
`sentence_transformers.losses.MSELoss.MSELoss`
Parameters of the fit()-Method:
```
{
"callback": null,
"epochs": 7,
"evaluation_steps": 1000,
"evaluator": "sentence_transformers.evaluation.SequentialEvaluator.SequentialEvaluator",
"max_grad_norm": 1,
"optimizer_class": "<class 'transformers.optimization.AdamW'>",
"optimizer_params": {
"correct_bias": false,
"eps": 1e-06,
"lr": 2e-05
},
"scheduler": "WarmupLinear",
"steps_per_epoch": null,
"warmup_steps": 10000,
"weight_decay": 0.01
}
```
## Full Model Architecture
```
SentenceTransformer(
(0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel
(1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)
```
## Citing & Authors
<!--- Describe where people can find more information -->
This model was trained by KBLab, a data lab at the National Library of Sweden.
You can cite the article on our blog: https://kb-labb.github.io/posts/2021-08-23-a-swedish-sentence-transformer/ .
```
@misc{rekathati2021introducing,
author = {Rekathati, Faton},
title = {The KBLab Blog: Introducing a Swedish Sentence Transformer},
url = {https://kb-labb.github.io/posts/2021-08-23-a-swedish-sentence-transformer/},
year = {2021}
}
```
## Acknowledgements
We gratefully acknowledge the HPC RIVR consortium ([www.hpc-rivr.si](https://www.hpc-rivr.si/)) and EuroHPC JU ([eurohpc-ju.europa.eu/](https://eurohpc-ju.europa.eu/)) for funding this research by providing computing resources of the HPC system Vega at the Institute of Information Science ([www.izum.si](https://www.izum.si/)). |
YituTech/conv-bert-small | 9a11330184f20d78feb3fd45edd1e8dad23205e8 | 2021-02-24T11:26:46.000Z | [
"pytorch",
"tf",
"convbert",
"feature-extraction",
"transformers"
] | feature-extraction | false | YituTech | null | YituTech/conv-bert-small | 851 | 1 | transformers | 1,890 | Entry not found |
microsoft/markuplm-base | b907337efa18696ad8a213005d5db0946d5d2081 | 2022-01-11T12:32:38.000Z | [
"pytorch",
"markuplm",
"arxiv:2110.08518",
"transformers"
] | null | false | microsoft | null | microsoft/markuplm-base | 851 | 2 | transformers | 1,891 | # MarkupLM
**Multimodal (text +markup language) pre-training for [Document AI](https://www.microsoft.com/en-us/research/project/document-ai/)**
## Introduction
MarkupLM is a simple but effective multi-modal pre-training method of text and markup language for visually-rich document understanding and information extraction tasks, such as webpage QA and webpage information extraction. MarkupLM archives the SOTA results on multiple datasets. For more details, please refer to our paper:
[MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) Junlong Li, Yiheng Xu, Lei Cui, Furu Wei
|
tuner007/pegasus_summarizer | a8980c11072794c107d4e8b7990c6a49f3da6a50 | 2022-07-28T06:38:07.000Z | [
"pytorch",
"pegasus",
"text2text-generation",
"en",
"transformers",
"seq2seq",
"summarization",
"license:apache-2.0",
"model-index",
"autotrain_compatible"
] | summarization | false | tuner007 | null | tuner007/pegasus_summarizer | 848 | 7 | transformers | 1,892 | ---
language: en
license: apache-2.0
tags:
- pegasus
- seq2seq
- summarization
model-index:
- name: tuner007/pegasus_summarizer
results:
- task:
type: summarization
name: Summarization
dataset:
name: cnn_dailymail
type: cnn_dailymail
config: 3.0.0
split: train
metrics:
- name: ROUGE-1
type: rouge
value: 36.604
verified: true
- name: ROUGE-2
type: rouge
value: 14.6398
verified: true
- name: ROUGE-L
type: rouge
value: 23.8845
verified: true
- name: ROUGE-LSUM
type: rouge
value: 32.9017
verified: true
- name: loss
type: loss
value: 2.5757133960723877
verified: true
- name: gen_len
type: gen_len
value: 76.3984
verified: true
---
## Model description
[PEGASUS](https://github.com/google-research/pegasus) fine-tuned for summarization
## Install "sentencepiece" library required for tokenizer
```
pip install sentencepiece
```
## Model in Action 🚀
```
import torch
from transformers import PegasusForConditionalGeneration, PegasusTokenizer
model_name = 'tuner007/pegasus_summarizer'
torch_device = 'cuda' if torch.cuda.is_available() else 'cpu'
tokenizer = PegasusTokenizer.from_pretrained(model_name)
model = PegasusForConditionalGeneration.from_pretrained(model_name).to(torch_device)
def get_response(input_text):
batch = tokenizer([input_text],truncation=True,padding='longest',max_length=1024, return_tensors="pt").to(torch_device)
gen_out = model.generate(**batch,max_length=128,num_beams=5, num_return_sequences=1, temperature=1.5)
output_text = tokenizer.batch_decode(gen_out, skip_special_tokens=True)
return output_text
```
#### Example:
context = """"
India wicket-keeper batsman Rishabh Pant has said someone from the crowd threw a ball on pacer Mohammed Siraj while he was fielding in the ongoing third Test against England on Wednesday. Pant revealed the incident made India skipper Virat Kohli "upset". "I think, somebody threw a ball inside, at Siraj, so he [Kohli] was upset," said Pant in a virtual press conference after the close of the first day\'s play."You can say whatever you want to chant, but don\'t throw things at the fielders and all those things. It is not good for cricket, I guess," he added.In the third session of the opening day of the third Test, a section of spectators seemed to have asked Siraj the score of the match to tease the pacer. The India pacer however came with a brilliant reply as he gestured 1-0 (India leading the Test series) towards the crowd.Earlier this month, during the second Test match, there was some bad crowd behaviour on a show as some unruly fans threw champagne corks at India batsman KL Rahul.Kohli also intervened and he was seen gesturing towards the opening batsman to know more about the incident. An over later, the TV visuals showed that many champagne corks were thrown inside the playing field, and the Indian players were visibly left frustrated.Coming back to the game, after bundling out India for 78, openers Rory Burns and Haseeb Hameed ensured that England took the honours on the opening day of the ongoing third Test.At stumps, England\'s score reads 120/0 and the hosts have extended their lead to 42 runs. For the Three Lions, Burns (52*) and Hameed (60*) are currently unbeaten at the crease.Talking about the pitch on opening day, Pant said, "They took the heavy roller, the wicket was much more settled down, and they batted nicely also," he said. "But when we batted, the wicket was slightly soft, and they bowled in good areas, but we could have applied [ourselves] much better."Both England batsmen managed to see off the final session and the hosts concluded the opening day with all ten wickets intact, extending the lead to 42.(ANI)
"""
```
get_response(context)
```
#### Output:
Team India wicketkeeper-batsman Rishabh Pant has said that Virat Kohli was "upset" after someone threw a ball on pacer Mohammed Siraj while he was fielding in the ongoing third Test against England. "You can say whatever you want to chant, but don't throw things at the fielders and all those things. It's not good for cricket, I guess," Pant added.'
#### [Inshort](https://www.inshorts.com/) (60 words News summary app, rated 4.4 by 5,27,246+ users on android playstore) summary:
India wicketkeeper-batsman Rishabh Pant has revealed that captain Virat Kohli was upset with the crowd during the first day of Leeds Test against England because someone threw a ball at pacer Mohammed Siraj. Pant added, "You can say whatever you want to chant, but don't throw things at the fielders and all those things. It is not good for cricket."
> Created by [Arpit Rajauria](https://twitter.com/arpit_rajauria)
[](https://twitter.com/arpit_rajauria)
|
vblagoje/retribert-base-uncased | 9241266f3afbc5b07435cfa8070871fc77ee3818 | 2021-11-11T07:23:38.000Z | [
"pytorch",
"retribert",
"feature-extraction",
"transformers"
] | feature-extraction | false | vblagoje | null | vblagoje/retribert-base-uncased | 847 | null | transformers | 1,893 | Entry not found |
ynie/xlnet-large-cased-snli_mnli_fever_anli_R1_R2_R3-nli | 027e2b37d8b0c27965ee58d9da95cf994f1ee0f4 | 2020-10-17T01:54:45.000Z | [
"pytorch",
"xlnet",
"text-classification",
"transformers"
] | text-classification | false | ynie | null | ynie/xlnet-large-cased-snli_mnli_fever_anli_R1_R2_R3-nli | 847 | null | transformers | 1,894 | Entry not found |
henryk/bert-base-multilingual-cased-finetuned-polish-squad2 | f4d5b523b23dbe9c3bc4741755832ef1451fe5ce | 2021-05-19T19:05:33.000Z | [
"pytorch",
"jax",
"bert",
"question-answering",
"pl",
"transformers",
"autotrain_compatible"
] | question-answering | false | henryk | null | henryk/bert-base-multilingual-cased-finetuned-polish-squad2 | 846 | 1 | transformers | 1,895 | ---
language: pl
---
# Multilingual + Polish SQuAD2.0
This model is the multilingual model provided by the Google research team with a fine-tuned polish Q&A downstream task.
## Details of the language model
Language model ([**bert-base-multilingual-cased**](https://github.com/google-research/bert/blob/master/multilingual.md)):
12-layer, 768-hidden, 12-heads, 110M parameters.
Trained on cased text in the top 104 languages with the largest Wikipedias.
## Details of the downstream task
Using the `mtranslate` Python module, [**SQuAD2.0**](https://rajpurkar.github.io/SQuAD-explorer/) was machine-translated. In order to find the start tokens, the direct translations of the answers were searched in the corresponding paragraphs. Due to the different translations depending on the context (missing context in the pure answer), the answer could not always be found in the text, and thus a loss of question-answer examples occurred. This is a potential problem where errors can occur in the data set.
| Dataset | # Q&A |
| ---------------------- | ----- |
| SQuAD2.0 Train | 130 K |
| Polish SQuAD2.0 Train | 83.1 K |
| SQuAD2.0 Dev | 12 K |
| Polish SQuAD2.0 Dev | 8.5 K |
## Model benchmark
| Model | EM/F1 |HasAns (EM/F1) | NoAns |
| ---------------------- | ----- | ----- | ----- |
| [SlavicBERT](https://huggingface.co/DeepPavlov/bert-base-bg-cs-pl-ru-cased) | 69.35/71.51 | 47.02/54.09 | 79.20 |
| [polBERT](https://huggingface.co/dkleczek/bert-base-polish-uncased-v1) | 67.33/69.80| 45.73/53.80 | 76.87 |
| [multiBERT](https://huggingface.co/bert-base-multilingual-cased) | **70.76**/**72.92** |45.00/52.04 | 82.13 |
## Model training
The model was trained on a **Tesla V100** GPU with the following command:
```python
export SQUAD_DIR=path/to/pl_squad
python run_squad.py
--model_type bert \
--model_name_or_path bert-base-multilingual-cased \
--do_train \
--do_eval \
--version_2_with_negative \
--train_file $SQUAD_DIR/pl_squadv2_train.json \
--predict_file $SQUAD_DIR/pl_squadv2_dev.json \
--num_train_epochs 2 \
--max_seq_length 384 \
--doc_stride 128 \
--save_steps=8000 \
--output_dir ../../output \
--overwrite_cache \
--overwrite_output_dir
```
**Results**:
{'exact': 70.76671723655035, 'f1': 72.92156947155917, 'total': 8569, 'HasAns_exact': 45.00762195121951, 'HasAns_f1': 52.04456128116991, 'HasAns_total': 2624, 'NoAns_exact': 82.13624894869638, '
NoAns_f1': 82.13624894869638, 'NoAns_total': 5945, 'best_exact': 71.72365503559342, 'best_exact_thresh': 0.0, 'best_f1': 73.62662512059369, 'best_f1_thresh': 0.0}
## Model in action
Fast usage with **pipelines**:
```python
from transformers import pipeline
qa_pipeline = pipeline(
"question-answering",
model="henryk/bert-base-multilingual-cased-finetuned-polish-squad2",
tokenizer="henryk/bert-base-multilingual-cased-finetuned-polish-squad2"
)
qa_pipeline({
'context': "Warszawa jest największym miastem w Polsce pod względem liczby ludności i powierzchni",
'question': "Jakie jest największe miasto w Polsce?"})
```
# Output:
```json
{
"score": 0.9986,
"start": 0,
"end": 8,
"answer": "Warszawa"
}
```
## Contact
Please do not hesitate to contact me via [LinkedIn](https://www.linkedin.com/in/henryk-borzymowski-0755a2167/) if you want to discuss or get access to the Polish version of SQuAD. |
LIAMF-USP/roberta-large-finetuned-race | 671db4772791326255cbf6c4f33eff5d06db4e43 | 2021-05-20T12:08:36.000Z | [
"pytorch",
"tf",
"jax",
"roberta",
"multiple-choice",
"english",
"dataset:race",
"transformers",
"license:mit"
] | multiple-choice | false | LIAMF-USP | null | LIAMF-USP/roberta-large-finetuned-race | 845 | 3 | transformers | 1,896 | ---
language: "english"
license: "mit"
datasets:
- race
metrics:
- accuracy
---
# Roberta Large Fine Tuned on RACE
## Model description
This model is a fine-tuned model of Roberta-large applied on RACE
#### How to use
```python
import datasets
from transformers import RobertaTokenizer
from transformers import RobertaForMultipleChoice
tokenizer = RobertaTokenizer.from_pretrained(
"LIAMF-USP/roberta-large-finetuned-race")
model = RobertaForMultipleChoice.from_pretrained(
"LIAMF-USP/roberta-large-finetuned-race")
dataset = datasets.load_dataset(
"race",
"all",
split=["train", "validation", "test"],
)training_examples = dataset[0]
evaluation_examples = dataset[1]
test_examples = dataset[2]
example=training_examples[0]
example_id = example["example_id"]
question = example["question"]
context = example["article"]
options = example["options"]
label_example = example["answer"]
label_map = {label: i
for i, label in enumerate(["A", "B", "C", "D"])}
choices_inputs = []
for ending_idx, (_, ending) in enumerate(
zip(context, options)):
if question.find("_") != -1:
# fill in the banks questions
question_option = question.replace("_", ending)
else:
question_option = question + " " + ending
inputs = tokenizer(
context,
question_option,
add_special_tokens=True,
max_length=MAX_SEQ_LENGTH,
padding="max_length",
truncation=True,
return_overflowing_tokens=False,
)
label = label_map[label_example]
input_ids = [x["input_ids"] for x in choices_inputs]
attention_mask = (
[x["attention_mask"] for x in choices_inputs]
# as the senteces follow the same structure,
#just one of them is necessary to check
if "attention_mask" in choices_inputs[0]
else None
)
example_encoded = {
"example_id": example_id,
"input_ids": input_ids,
"attention_mask": attention_mask,
"label": label,
}
output = model(**example_encoded)
```
## Training data
The initial model was [roberta large model](https://huggingface.co/roberta-large) which was then fine-tuned on [RACE dataset](https://www.cs.cmu.edu/~glai1/data/race/)
## Training procedure
It was necessary to preprocess the data with a method that is exemplified for a single instance in the _How to use_ section. The used hyperparameters were the following:
| Hyperparameter | Value |
|:----:|:----:|
| adam_beta1 | 0.9 |
| adam_beta2 | 0.98 |
| adam_epsilon | 1.000e-8 |
| eval_batch_size | 32 |
| train_batch_size | 1 |
| fp16 | True |
| gradient_accumulation_steps | 16 |
| learning_rate | 0.00001 |
| warmup_steps | 1000 |
| max_length | 512 |
| epochs | 4 |
## Eval results:
| Dataset Acc | Eval | All Test |High School Test |Middle School Test |
|:----:|:----:|:----:|:----:|:----:|
| | 85.2 | 84.9|83.5|88.0|
**The model was trained with a Tesla V100-PCIE-16GB** |
microsoft/xprophetnet-large-wiki100-cased | 1acad1643ddd54a44df6a1b797ada8373685d90e | 2020-12-11T21:51:18.000Z | [
"pytorch",
"xlm-prophetnet",
"text2text-generation",
"multilingual",
"arxiv:2001.04063",
"arxiv:2004.01401",
"transformers",
"autotrain_compatible"
] | text2text-generation | false | microsoft | null | microsoft/xprophetnet-large-wiki100-cased | 845 | null | transformers | 1,897 | ---
language: multilingual
---
## xprophetnet-large-wiki100-cased
Cross-lingual version [ProphetNet](https://arxiv.org/abs/2001.04063), pretrained on [wiki100 xGLUE dataset](https://arxiv.org/abs/2004.01401).
ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.
ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).
xProphetNet is also served as the baseline model for xGLUE cross-lingual natural language generation tasks.
For xGLUE corss-lingual NLG tasks, xProphetNet is finetuned with English data, but inference with both English and other zero-shot language data.
### Usage
This pre-trained model can be fine-tuned on *sequence-to-sequence* tasks. The model could *e.g.* be trained on English headline generation as follows:
```python
from transformers import XLMProphetNetForConditionalGeneration, XLMProphetNetTokenizer
model = XLMProphetNetForConditionalGeneration.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
tokenizer = XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
input_str = "the us state department said wednesday it had received no formal word from bolivia that it was expelling the us ambassador there but said the charges made against him are `` baseless ."
target_str = "us rejects charges against its ambassador in bolivia"
input_ids = tokenizer(input_str, return_tensors="pt").input_ids
labels = tokenizer(target_str, return_tensors="pt").input_ids
loss = model(input_ids, labels=labels).loss
```
Note that since this model is a multi-lingual model it can be fine-tuned on all kinds of other languages.
### Citation
```bibtex
@article{yan2020prophetnet,
title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
journal={arXiv preprint arXiv:2001.04063},
year={2020}
}
```
|
kuzgunlar/electra-turkish-qa | 586ab1bc12af16bf396360db8a90cdccf514e4ec | 2020-07-31T09:15:54.000Z | [
"pytorch",
"electra",
"question-answering",
"transformers",
"autotrain_compatible"
] | question-answering | false | kuzgunlar | null | kuzgunlar/electra-turkish-qa | 843 | 1 | transformers | 1,898 | Entry not found |
IDEA-CCNL/Erlangshen-Roberta-110M-NLI | ea0a42559acf675d8931336951e893fc5d466268 | 2022-05-12T09:48:51.000Z | [
"pytorch",
"bert",
"text-classification",
"zh",
"transformers",
"NLU",
"NLI",
"license:apache-2.0"
] | text-classification | false | IDEA-CCNL | null | IDEA-CCNL/Erlangshen-Roberta-110M-NLI | 843 | null | transformers | 1,899 | ---
language:
- zh
license: apache-2.0
tags:
- bert
- NLU
- NLI
inference: true
widget:
- text: "今天心情不好[SEP]今天很开心"
---
# Erlangshen-Roberta-110M-NLI, model (Chinese)，one model of [Fengshenbang-LM](https://github.com/IDEA-CCNL/Fengshenbang-LM).
We collect 4 NLI（Natural Language Inference） datasets in the Chinese domain for finetune, with a total of 1014787 samples. Our model is mainly based on [roberta](https://huggingface.co/hfl/chinese-roberta-wwm-ext)
## Usage
```python
from transformers import BertForSequenceClassification
from transformers import BertTokenizer
import torch
tokenizer=BertTokenizer.from_pretrained('IDEA-CCNL/Erlangshen-Roberta-110M-NLI')
model=BertForSequenceClassification.from_pretrained('IDEA-CCNL/Erlangshen-Roberta-110M-NLI')
texta='今天的饭不好吃'
textb='今天心情不好'
output=model(torch.tensor([tokenizer.encode(texta,textb)]))
print(torch.nn.functional.softmax(output.logits,dim=-1))
```
## Scores on downstream chinese tasks (without any data augmentation)
| Model | cmnli | ocnli | snli |
| :--------: | :-----: | :----: | :-----: |
| Erlangshen-Roberta-110M-NLI | 80.83 | 78.56 | 88.01 |
| Erlangshen-Roberta-330M-NLI | 82.25 | 79.82 | 88 |
| Erlangshen-MegatronBert-1.3B-NLI | 84.52 | 84.17 | 88.67 |
## Citation
If you find the resource is useful, please cite the following website in your paper.
```
@misc{Fengshenbang-LM,
title={Fengshenbang-LM},
author={IDEA-CCNL},
year={2021},
howpublished={\url{https://github.com/IDEA-CCNL/Fengshenbang-LM}},
}
``` |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.