Update README.md
Browse files
README.md
CHANGED
|
@@ -1,9 +1,64 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-nd-4.0
|
| 3 |
---
|
| 4 |
-
Mistral & code dataset finetuning qlora
|
| 5 |
|
| 6 |
-
|
| 7 |
|
| 8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-nd-4.0
|
| 3 |
---
|
|
|
|
| 4 |
|
| 5 |
+
# Mistral-7B-Instruct-v0.2-code-ft
|
| 6 |
|
| 7 |
+
I'm thrilled to introduce the latest iteration of our model, Mistral-7B-Instruct-v0.2-code-ft. This updated version is designed to further enhance coding assistance and co-pilot functionalities. We're eager for developers and enthusiasts to try it out and provide feedback!
|
| 8 |
+
|
| 9 |
+
## Additional Information
|
| 10 |
+
|
| 11 |
+
This version builds upon the previous Mistral-7B models, incorporating new datasets and features for a more refined experience.
|
| 12 |
+
|
| 13 |
+
## Prompt template: ChatML
|
| 14 |
+
```
|
| 15 |
+
<|im_start|>system
|
| 16 |
+
{system_message}<|im_end|>
|
| 17 |
+
<|im_start|>user
|
| 18 |
+
{prompt}<|im_end|>
|
| 19 |
+
<|im_start|>assistant
|
| 20 |
+
```
|
| 21 |
+
|
| 22 |
+
## Eval Plus Performance
|
| 23 |
+
|
| 24 |
+
For detailed performance metrics, visit Eval Plus page: [Mistral-7B-Instruct-v0.2-code-ft Eval Plus](https://github.com/evalplus/evalplus)
|
| 25 |
+
|
| 26 |
+
Score: 0.421
|
| 27 |
|
| 28 |
+
|
| 29 |
+
## Dataset:
|
| 30 |
+
The model has been trained on a new dataset to improve its performance and versatility:
|
| 31 |
+
- path: ajibawa-2023/Code-74k-ShareGPT
|
| 32 |
+
|
| 33 |
+
type: sharegpt
|
| 34 |
+
|
| 35 |
+
conversation: chatml
|
| 36 |
+
|
| 37 |
+
Find more about the dataset here: [Code-74k-ShareGPT Dataset](https://huggingface.co/datasets/ajibawa-2023/Code-74k-ShareGPT)
|
| 38 |
+
|
| 39 |
+
## Model Architecture
|
| 40 |
+
|
| 41 |
+
- Base Model: mistralai/Mistral-7B-Instruct-v0.2
|
| 42 |
+
- Tokenizer Type: LlamaTokenizer
|
| 43 |
+
- Model Type: MistralForCausalLM
|
| 44 |
+
- Is Mistral Derived Model: true
|
| 45 |
+
- Sequence Length: 16384 with sample packing
|
| 46 |
+
|
| 47 |
+
## Enhanced Features
|
| 48 |
+
|
| 49 |
+
- Adapter: qlora
|
| 50 |
+
- Learning Rate: 0.0002 with cosine lr scheduler
|
| 51 |
+
- Optimizer: paged_adamw_32bit
|
| 52 |
+
- Training Enhancements: bf16 training, gradient checkpointing, and flash attention
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
## Download Information
|
| 56 |
+
|
| 57 |
+
You can download and explore this model through these links on Hugging Face.
|
| 58 |
+
|
| 59 |
+
## Contributions and Feedback
|
| 60 |
+
|
| 61 |
+
We welcome contributions and feedback from the community. Please feel free to open issues or pull requests on repository.
|
| 62 |
+
|
| 63 |
+
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|
| 64 |
+
|