README.md

---
library_name: transformers
license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
- axolotl
- generated_from_trainer
model-index:
- name: Mistral-7B-v0.1-q-sparse-fineweb-edu-table2-re
  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. -->

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.5.2`
```yaml
base_model: mistralai/Mistral-7B-v0.1
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
tokenizer_use_fast: false
resize_token_embeddings_to_32x: false

flash_attention: true
xformers_attention:

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: skymizer/Mistral-7B-v0.1-base-tokenized-fineweb-edu-45B-4096
    train_on_split: train
    type: completion

test_datasets:
  - path: skymizer/Mistral-7B-v0.1-base-tokenized-fineweb-edu-test-4K
    split: test
    type: completion

is_preprocess: true
skip_prepare_dataset: true

dataset_prepared_path:

hf_use_auth_token: true
output_dir: /mnt/home/model-team/models/Mistral-7B-v0.1-q-sparse-fineweb-edu-table2-re
resume_from_checkpoint:
auto_resume_from_checkpoints: true

sequence_len: 4096
sample_packing: true
sample_packing_group_size: 100000
sample_packing_bin_size: 200
pad_to_sequence_len: true

eval_sample_packing: false
# eval_causal_lm_metrics: ["perplexity"]

wandb_project: "sparse-tuning-cpt"
wandb_entity:
wandb_watch:
wandb_name: "Mistral-7B-v0.1-q-sparse-fineweb-edu-table2-re"
wandb_log_model:

# global batch size = 2 * 8 * 8 GPUs * 8 Nodes * 4096 = 4M
gradient_accumulation_steps: 2
micro_batch_size: 8
eval_batch_size: 1
max_steps: 10000
optimizer: adamw_torch
learning_rate: 0.00005
lr_scheduler: cosine
cosine_min_lr_ratio: 0.2 
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.95
adam_eps: 0.000001
max_grad_norm: 2.0

train_on_inputs: false
group_by_length: false
bf16: true
fp16:
tf32: false

hub_model_id: "skymizer/Mistral-7B-v0.1-q-sparse-fineweb-edu-table2-re"

save_strategy: "steps"
save_steps: 500

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
local_rank:
logging_steps: 1

warmup_steps: 375
eval_steps: 500
eval_table_size:
debug:
deepspeed: /root/train/axolotl/deepspeed_configs/zero3_bf16.json
fsdp:
fsdp_config:
seed: 42

```

</details><br>

# Mistral-7B-v0.1-q-sparse-fineweb-edu-table2-re

This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 1.9784

## 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: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 64
- gradient_accumulation_steps: 2
- total_train_batch_size: 1024
- total_eval_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 375
- training_steps: 10000

### Training results

| Training Loss | Epoch  | Step  | Validation Loss |
|:-------------:|:------:|:-----:|:---------------:|
| 11.1526       | 0.0001 | 1     | 11.1178         |
| 3.9513        | 0.0408 | 500   | 3.7699          |
| 3.4469        | 0.0817 | 1000  | 3.2772          |
| 3.1993        | 0.1225 | 1500  | 3.0024          |
| 2.8081        | 0.1633 | 2000  | 2.7218          |
| 2.5217        | 0.2042 | 2500  | 2.4860          |
| 2.3993        | 0.2450 | 3000  | 2.3570          |
| 2.2919        | 0.2858 | 3500  | 2.2761          |
| 2.2379        | 0.3267 | 4000  | 2.2180          |
| 2.2047        | 0.3675 | 4500  | 2.1721          |
| 2.1553        | 0.4083 | 5000  | 2.1367          |
| 2.1279        | 0.4491 | 5500  | 2.1066          |
| 2.0689        | 0.4900 | 6000  | 2.0822          |
| 2.0702        | 0.5308 | 6500  | 2.0608          |
| 2.0611        | 0.5716 | 7000  | 2.0425          |
| 2.0242        | 0.6125 | 7500  | 2.0264          |
| 2.0449        | 0.6533 | 8000  | 2.0140          |
| 2.0245        | 0.6941 | 8500  | 2.0025          |
| 2.0107        | 0.7350 | 9000  | 1.9933          |
| 1.9995        | 0.7758 | 9500  | 1.9851          |
| 1.9995        | 0.8166 | 10000 | 1.9784          |


### Framework versions

- Transformers 4.46.3
- Pytorch 2.5.1+cu124
- Datasets 3.1.0
- Tokenizers 0.20.3