|
--- |
|
license: mit |
|
base_model: mistralai/Mistral-7B-v0.1 |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: supercot-lora |
|
results: [] |
|
--- |
|
|
|
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl) |
|
# mistral-v0.1-supercot-lora |
|
|
|
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the [supercot](https://huggingface.co/datasets/kaiokendev/SuperCOT-dataset) dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.9790 |
|
|
|
## Model description |
|
|
|
SuperCOT is a LoRA trained with the aim of making Mistral follow prompts for Langchain better, by infusing chain-of-thought datasets, code explanations and instructions, snippets, logical deductions and Alpaca GPT-4 prompts. It uses a mixture of the following datasets: |
|
|
|
https://huggingface.co/datasets/QingyiSi/Alpaca-CoT |
|
- Chain of thought QED |
|
- Chain of thought Aqua |
|
- CodeAlpaca |
|
|
|
https://huggingface.co/datasets/neulab/conala |
|
- Code snippets |
|
|
|
https://huggingface.co/datasets/yahma/alpaca-cleaned |
|
- Alpaca GPT4 |
|
|
|
## Intended uses & limitations |
|
|
|
The model will show biases similar to those exhibited by the base model. It is not intended for supplying factual information or advice in any form. |
|
|
|
## Training and evaluation data |
|
|
|
[kaiokendev/SuperCOT-dataset](https://huggingface.co/datasets/kaiokendev/SuperCOT-dataset) |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0002 |
|
- train_batch_size: 2 |
|
- eval_batch_size: 2 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 4 |
|
- total_train_batch_size: 8 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_steps: 10 |
|
- num_epochs: 3 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|:-------------:|:-----:|:----:|:---------------:| |
|
| 0.7661 | 0.06 | 20 | 1.5173 | |
|
| 0.7681 | 0.12 | 40 | 1.2323 | |
|
| 0.6647 | 0.18 | 60 | 1.1306 | |
|
| 0.6742 | 0.24 | 80 | 1.0847 | |
|
| 0.6995 | 0.3 | 100 | 1.0573 | |
|
| 0.6883 | 0.36 | 120 | 1.0412 | |
|
| 0.6437 | 0.42 | 140 | 1.0375 | |
|
| 0.6331 | 0.48 | 160 | 1.0186 | |
|
| 0.6686 | 0.54 | 180 | 1.0153 | |
|
| 0.6767 | 0.6 | 200 | 1.0042 | |
|
| 0.7037 | 0.66 | 220 | 1.0023 | |
|
| 0.6994 | 0.72 | 240 | 1.0014 | |
|
| 0.7012 | 0.78 | 260 | 0.9996 | |
|
| 0.6599 | 0.84 | 280 | 0.9926 | |
|
| 0.6401 | 0.9 | 300 | 0.9913 | |
|
| 0.6665 | 0.96 | 320 | 0.9910 | |
|
| 0.5771 | 1.02 | 340 | 0.9907 | |
|
| 0.6286 | 1.08 | 360 | 0.9830 | |
|
| 0.6064 | 1.14 | 380 | 0.9865 | |
|
| 0.5976 | 1.19 | 400 | 0.9802 | |
|
| 0.5512 | 1.25 | 420 | 0.9817 | |
|
| 0.6333 | 1.31 | 440 | 0.9810 | |
|
| 0.5883 | 1.37 | 460 | 0.9817 | |
|
| 0.5822 | 1.43 | 480 | 0.9783 | |
|
| 0.5878 | 1.49 | 500 | 0.9757 | |
|
| 0.5951 | 1.55 | 520 | 0.9753 | |
|
| 0.6466 | 1.61 | 540 | 0.9719 | |
|
| 0.6246 | 1.67 | 560 | 0.9681 | |
|
| 0.627 | 1.73 | 580 | 0.9705 | |
|
| 0.6214 | 1.79 | 600 | 0.9691 | |
|
| 0.6558 | 1.85 | 620 | 0.9709 | |
|
| 0.5736 | 1.91 | 640 | 0.9674 | |
|
| 0.6188 | 1.97 | 660 | 0.9674 | |
|
| 0.5293 | 2.03 | 680 | 0.9742 | |
|
| 0.5463 | 2.09 | 700 | 0.9766 | |
|
| 0.5184 | 2.15 | 720 | 0.9776 | |
|
| 0.5349 | 2.21 | 740 | 0.9783 | |
|
| 0.5536 | 2.27 | 760 | 0.9794 | |
|
| 0.5016 | 2.33 | 780 | 0.9822 | |
|
| 0.5075 | 2.39 | 800 | 0.9795 | |
|
| 0.5529 | 2.45 | 820 | 0.9808 | |
|
| 0.5168 | 2.51 | 840 | 0.9784 | |
|
| 0.5416 | 2.57 | 860 | 0.9793 | |
|
| 0.4845 | 2.63 | 880 | 0.9804 | |
|
| 0.5487 | 2.69 | 900 | 0.9801 | |
|
| 0.5313 | 2.75 | 920 | 0.9797 | |
|
| 0.5449 | 2.81 | 940 | 0.9790 | |
|
| 0.5303 | 2.87 | 960 | 0.9795 | |
|
| 0.5599 | 2.93 | 980 | 0.9795 | |
|
| 0.544 | 2.99 | 1000 | 0.9790 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.34.0.dev0 |
|
- Pytorch 2.0.1+cu118 |
|
- Datasets 2.14.5 |
|
- Tokenizers 0.14.0 |
|
|
|
### Citations |
|
|
|
Alpaca COT datasets |
|
``` |
|
@misc{alpaca-cot, |
|
author = {Qingyi Si, Zheng Lin }, |
|
school = {Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China}, |
|
title = {Alpaca-CoT: An Instruction Fine-Tuning Platform with Instruction Data Collection and Unified Large Language Models Interface}, |
|
year = {2023}, |
|
publisher = {GitHub}, |
|
journal = {GitHub repository}, |
|
howpublished = {\url{https://github.com/PhoebusSi/alpaca-CoT}}, |
|
} |
|
``` |
|
Stanford Alpaca |
|
``` |
|
@misc{alpaca, |
|
author = {Rohan Taori and Ishaan Gulrajani and Tianyi Zhang and Yann Dubois and Xuechen Li and Carlos Guestrin and Percy Liang and Tatsunori B. Hashimoto }, |
|
title = {Stanford Alpaca: An Instruction-following LLaMA model}, |
|
year = {2023}, |
|
publisher = {GitHub}, |
|
journal = {GitHub repository}, |
|
howpublished = {\url{https://github.com/tatsu-lab/stanford_alpaca}}, |
|
} |
|
``` |
|
Google FLAN |
|
``` |
|
@inproceedings{weifinetuned, |
|
title={Finetuned Language Models are Zero-Shot Learners}, |
|
author={Wei, Jason and Bosma, Maarten and Zhao, Vincent and Guu, Kelvin and Yu, Adams Wei and Lester, Brian and Du, Nan and Dai, Andrew M and Le, Quoc V}, |
|
booktitle={International Conference on Learning Representations} |
|
} |
|
``` |