Update README.md
Browse files
README.md
CHANGED
@@ -10,10 +10,44 @@ datasets:
|
|
10 |
- Anthropic/hh-rlhf
|
11 |
---
|
12 |
|
|
|
|
|
13 |
Pythia-1b supervised finetuned with Anthropic-hh-rlhf dataset for 1 epoch.
|
14 |
|
15 |
[wandb log](https://wandb.ai/pythia_dpo/Pythia_DPO_new/runs/xk2ub7ig?workspace=user-leogrin)
|
16 |
|
17 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
|
19 |
-
See [Pythia-1b](https://huggingface.co/EleutherAI/pythia-1b) for model details [(paper)](https://arxiv.org/abs/2101.00027).
|
|
|
10 |
- Anthropic/hh-rlhf
|
11 |
---
|
12 |
|
13 |
+
# Infos
|
14 |
+
|
15 |
Pythia-1b supervised finetuned with Anthropic-hh-rlhf dataset for 1 epoch.
|
16 |
|
17 |
[wandb log](https://wandb.ai/pythia_dpo/Pythia_DPO_new/runs/xk2ub7ig?workspace=user-leogrin)
|
18 |
|
19 |
+
See [Pythia-1b](https://huggingface.co/EleutherAI/pythia-1b) for model details [(paper)](https://arxiv.org/abs/2101.00027).
|
20 |
+
|
21 |
+
|
22 |
+
# Benchmark results:
|
23 |
+
|
24 |
+
## Zero shot
|
25 |
+
|
26 |
+
Results for the base model are taken from the [Pythia paper](https://arxiv.org/abs/2101.00027).
|
27 |
+
|
28 |
+
| Task | 1B_base | 1B_sft |
|
29 |
+
|------------------|----------------|----------------|
|
30 |
+
| Lambada (OpenAI) | 0.562 ± 0.007 | 0.563 ± 0.007 |
|
31 |
+
| PIQA | 0.707 ± 0.011 | 0.711 ± 0.011 |
|
32 |
+
| WinoGrande | 0.537 ± 0.014 | 0.534 ± 0.014 |
|
33 |
+
| WSC | 0.365 ± 0.047 | 0.365 ± 0.047 |
|
34 |
+
| ARC - Easy | 0.569 ± 0.010 | 0.583 ± 0.010 |
|
35 |
+
| ARC - Challenge | 0.244 ± 0.013 | 0.248 ± 0.013 |
|
36 |
+
| SciQ | 0.840 ± 0.012 | 0.847 ± 0.011 |
|
37 |
+
| LogiQA | 0.223 ± 0.016 | -- |
|
38 |
+
|
39 |
+
## Five shot
|
40 |
+
|
41 |
+
| Task | 1B_base | 1B_sft |
|
42 |
+
|------------------|----------------|----------------|
|
43 |
+
| Lambada (OpenAI) | 0.507 ± 0.007 | 0.4722 ± 0.007 |
|
44 |
+
| PIQA | 0.705 ± 0.011 | 0.7165 ± 0.0105|
|
45 |
+
| WinoGrande | 0.532 ± 0.014 | 0.5343 ± 0.014 |
|
46 |
+
| WSC | 0.365 ± 0.047 | 0.5000 ± 0.0493|
|
47 |
+
| ARC - Easy | 0.594 ± 0.010 | 0.6010 ± 0.010 |
|
48 |
+
| ARC - Challenge | 0.259 ± 0.013 | 0.2679 ± 0.0129|
|
49 |
+
| SciQ | 0.920 ± 0.009 | 0.9100 ± 0.0091|
|
50 |
+
| LogiQA | 0.227 ± 0.016 | N/A |
|
51 |
+
|
52 |
+
|
53 |
|
|