Commit
·
a6d13ee
1
Parent(s):
a36cb6f
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,16 +1,24 @@
|
|
| 1 |
---
|
| 2 |
license: openrail
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
It outperforms GPT-4 and ChatGPT (paper link soon) on a derivation generation task in ROUGE, BLEU, BLEURT, and GLEU, and shows some generalisation capabilities.
|
| 8 |
|
|
|
|
|
|
|
| 9 |
It was trained on 155 physics symbols, but struggles with out-of-vocabulary symbols.
|
| 10 |
|
| 11 |
|
| 12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: openrail
|
| 3 |
+
pipeline_tag: text-generation
|
| 4 |
---
|
| 5 |
|
| 6 |
+
**Overview**
|
|
|
|
|
|
|
| 7 |
|
| 8 |
+
MathT5-large is a version of FLAN-T5-large fine-tuned for 25 epochs on 15K (LaTeX) synthetic mathematical derivations (containing 5 - 9 equations), that were generated using a symbolic solver (SymPy).
|
| 9 |
+
It outperforms GPT-4 and ChatGPT (paper link soon) on a derivation generation task in ROUGE, BLEU, BLEURT, and GLEU scores, and shows some generalisation capabilities.
|
| 10 |
It was trained on 155 physics symbols, but struggles with out-of-vocabulary symbols.
|
| 11 |
|
| 12 |
|
| 13 |
+
**Example prompt:**
|
| 14 |
+
|
| 15 |
+
```prompt = "Given \\cos{(q)} = \\theta{(q)},
|
| 16 |
+
then derive - \\sin{(q)} = \\frac{d}{d q} \\theta{(q)},
|
| 17 |
+
then obtain (- \\sin{(q)})^{q} (\\frac{d}{d q} \\cos{(q)})^{q} = (- \\sin{(q)})^{2 q}"```
|
| 18 |
+
|
| 19 |
+
**To use MathT5 easily:**
|
| 20 |
|
| 21 |
+
1. Download ```MathT5.py``` to your working directory.
|
| 22 |
+
2. ```from MathT5 import load_model, inference```
|
| 23 |
+
3. ```tokenizer, model = load_model("jmeadows17/MathT5-large")```
|
| 24 |
+
4. ```inference(prompt, tokenizer, model)```
|