Update README.md
#1
by
project-baize
- opened
README.md
CHANGED
|
@@ -1,3 +1,31 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-nc-4.0
|
| 3 |
---
|
| 4 |
+
<p align="center">
|
| 5 |
+
<img width="500px" alt="Project Baize" src="https://user-images.githubusercontent.com/22514219/229195563-0cddfa74-e52f-4413-b4b4-e4ba489c4b3d.png">
|
| 6 |
+
</p>
|
| 7 |
+
<hr>
|
| 8 |
+
|
| 9 |
+
## What's Baize?
|
| 10 |
+
Baize is an open-source chat model fine-tuned with [LoRA](https://github.com/microsoft/LoRA). It uses 100k dialogs generated by letting ChatGPT chat with itself. We also use Alpaca's data to improve its performance. This repo contains 7B model.
|
| 11 |
+
|
| 12 |
+
## Why it's called Baize?
|
| 13 |
+
Baize (白泽) is a mythical creature in Chinese folklore, who speaks human languages and knows everything. This is exactly what we expect from a chat model.
|
| 14 |
+
|
| 15 |
+
## Training Parameters
|
| 16 |
+
|
| 17 |
+
- Base Model: [decapoda-research/llama-7b-hf](https://huggingface.co/decapoda-research/llama-7b-hf)
|
| 18 |
+
- Training Epoch: 1
|
| 19 |
+
- Batch Size: 64
|
| 20 |
+
- Maximum Input Length: 512
|
| 21 |
+
- Learning Rate: 2e-4
|
| 22 |
+
- LoRA Rank: 8
|
| 23 |
+
- Updated Modules: All Linears
|
| 24 |
+
|
| 25 |
+
## Training Dataset
|
| 26 |
+
|
| 27 |
+
- [Standford Alpaca](https://github.com/tatsu-lab/stanford_alpaca) (51,942)
|
| 28 |
+
- [Quora Dialogs](https://github.com/project-baize/baize) (54,456):
|
| 29 |
+
- [StackOverflow Dialogs](https://github.com/project-baize/baize) (57,046)
|
| 30 |
+
|
| 31 |
+
More details can be found in the Baize [GitHub]((https://github.com/project-baize/baize))
|