|
--- |
|
language: |
|
- en |
|
tags: |
|
- pytorch |
|
- causal-lm |
|
- pythia |
|
license: apache-2.0 |
|
datasets: |
|
- Anthropic/hh-rlhf |
|
--- |
|
|
|
# Infos |
|
|
|
Pythia-1.4b supervised finetuned with Anthropic-hh-rlhf dataset for 1 epoch. |
|
|
|
[wandb log](https://wandb.ai/pythia_dpo/Pythia_DPO_new/runs/xm0pxfej) |
|
|
|
See [Pythia-1.4b](https://huggingface.co/EleutherAI/pythia-1.4b) for model details [(paper)](https://arxiv.org/abs/2101.00027). |
|
|
|
|
|
# Benchmark raw results: |
|
|
|
Results for the base model are taken from the [Pythia paper](https://arxiv.org/abs/2101.00027). |
|
|
|
|
|
## Zero shot |
|
|
|
| Task | 1.4B_base | 1.4B_sft | |
|
|------------------|--------------:|--------------:| |
|
| Lambada (OpenAI) | 0.616 ± 0.007 | 0.5977 ± 0.0068 | |
|
| PIQA | 0.711 ± 0.011 | 0.7133 ± 0.0106 | |
|
| WinoGrande | 0.573 ± 0.014 | 0.5793 ± 0.0139 | |
|
| WSC | 0.365 ± 0.047 | 0.3654 ± 0.0474 | |
|
| ARC - Easy | 0.606 ± 0.010 | 0.6098 ± 0.0100 | |
|
| ARC - Challenge | 0.260 ± 0.013 | 0.2696 ± 0.0130 | |
|
| SciQ | 0.865 ± 0.011 | 0.8540 ± 0.0112 | |
|
| LogiQA | 0.210 ± 0.016 | NA | |
|
|
|
|
|
## Five shot |
|
|
|
| Task | 1.4B_base | 1.4B_sft | |
|
|------------------|----------------:|----------------:| |
|
| Lambada (OpenAI) | 0.578 ± 0.007 | 0.5201 ± 0.007 | |
|
| PIQA | 0.705 ± 0.011 | 0.7176 ± 0.0105| |
|
| WinoGrande | 0.580 ± 0.014 | 0.5793 ± 0.0139| |
|
| WSC | 0.365 ± 0.047 | 0.5288 ± 0.0492| |
|
| ARC - Easy | 0.643 ± 0.010 | 0.6376 ± 0.0099| |
|
| ARC - Challenge | 0.290 ± 0.013 | 0.2935 ± 0.0133| |
|
| SciQ | 0.92 ± 0.009 | 0.9180 ± 0.0087| |
|
| LogiQA | 0.240 ± 0.017 | N/A | |
|
|
|
|
|
|
|
|
|
|
