Commit
•
007c15e
1
Parent(s):
56c42ee
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,39 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- gair-prox/open-web-math-pro
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
base_model:
|
| 8 |
+
- TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
# TinyLlama-1.1B-ProXMath
|
| 14 |
+
|
| 15 |
+
<p align="center">
|
| 16 |
+
<img src="prox-teaser.png">
|
| 17 |
+
</p>
|
| 18 |
+
|
| 19 |
+
[ArXiv](http://arxiv.org/abs/xxxx) | [Data: OpenWebMath-Pro](https://huggingface.co/datasets/gair-prox/open-web-math-pro) | [Code](https://github.com/GAIR-NLP/program-every-example)
|
| 20 |
+
|
| 21 |
+
**TinyLlama-1.1B-ProXMath** is a math-adapted TinyLlama-1.1B model that is continually pre-trained on [OpenWebMath-Pro](https://huggingface.co/datasets/gair-prox/open-web-math-pro) (a refined version by ProX) for **15**B tokens.
|
| 22 |
+
|
| 23 |
+
## Evaluations
|
| 24 |
+
|
| 25 |
+
ProX models are evaluated on 9 common math reasoning benchmarks.
|
| 26 |
+
|
| 27 |
+
| Model | asdiv | gsm8k | mathqa | mawps | minerva_math | mmlu_stem | sat_math | svamp | tabmwp | average |
|
| 28 |
+
|-------------------------|:--------:|:-------:|:--------:|:--------:|:------------:|:---------:|:--------:|:--------:|:--------:|:--------:|
|
| 29 |
+
| TinyLlama-1.1B | 18.0 | 2.8 | 14.6 | 20.2 | 3.2 | 16.3 | 21.9 | 10.9 | 12.5 | 13.4 |
|
| 30 |
+
| TinyLlama-1.1B-ProXMath | **41.9** | **9.0** | **15.6** | **56.9** | **5.6** | **26.8** | **31.2** | **23.8** | **22.2** | **25.7** |
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
### Citation
|
| 34 |
+
```
|
| 35 |
+
@misc{TBD
|
| 36 |
+
}
|
| 37 |
+
```
|
| 38 |
+
|
| 39 |
+
|