Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,31 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
model-index:
|
3 |
+
- name: MetaMath-LoRA-LLaMA-7B
|
4 |
+
results:
|
5 |
+
- task:
|
6 |
+
type: text-generation
|
7 |
+
dataset:
|
8 |
+
name: meta-math/MetaMathQA
|
9 |
+
type: meta-math/MetaMathQA
|
10 |
+
metrics:
|
11 |
+
- name: Accuracy (zero-shot)
|
12 |
+
type: Accuracy (zero-shot)
|
13 |
+
value: 0.61
|
14 |
+
verified: true
|
15 |
+
source:
|
16 |
+
name: Arithmetic Reasoning on GSM8K
|
17 |
+
url: https://paperswithcode.com/sota/arithmetic-reasoning-on-gsm8k
|
18 |
license: mit
|
19 |
---
|
20 |
+
|
21 |
+
# Fine-tune LLaMA 2 (7B) with LoRA on meta-math/MetaMathQA
|
22 |
+
Fine-tune for two epochs
|
23 |
+
|
24 |
+
## Result:
|
25 |
+
**Reload the saved adapter**:
|
26 |
+
Invalid output length: 9, Testing length: 1319, **Accuracy: **
|
27 |
+
|
28 |
+
## Comparison
|
29 |
+
The official report **accuracy is 0.665** by fine-tuning the whole LLaMA 2 7B model for 3 epochs.
|
30 |
+
|
31 |
+
**Note**: The LoRA adapter is being used for future research purposes.
|