Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[P1] Addressing issues in DPO Training (#127) #128

Merged
merged 1 commit into from
Aug 12, 2024
Merged

[P1] Addressing issues in DPO Training (#127) #128

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion examples/dpo/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,4 +4,6 @@ This is a tutorial for using ReFT with the [Direct Preference Optimization (DPO)

Follow the [`dpo.ipynb`](dpo.ipynb) notebook for a walk-through of training a ReFT model with DPO to answer questions truthfully based on the [TruthfulQA](https://arxiv.org/abs/2109.07958) dataset.

The DPO ReFT trainer is based on the DPOTrainer implementation in the `trl` library. The adapted trainer is implemented in [`dpo_trainer.py`](dpo_trainer.py).
The DPO ReFT trainer is based on the DPOTrainer implementation in the `trl` library (note: please use v0.8.6). The adapted trainer is implemented in [`dpo_trainer.py`](dpo_trainer.py).

We also provide a python script, [`dpo.py`](dpo.py), which you can run instead of the notebook tutorial. To see the W&B logs from running `dpo.py` with default parameters, see the [project here](https://wandb.ai/amirzur1212/reft_dpo/workspace).
Loading