-
Notifications
You must be signed in to change notification settings - Fork 404
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.
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
readme updates for PPO and Qwen2 #1267
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/1267
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 6ac9147 with merge base c593c10 (): This comment was automatically generated by Dr. CI and updates every 15 minutes. |
@@ -306,6 +306,8 @@ torchtune provides well-tested components with a high-bar on correctness. The li | |||
|
|||
We really value our community and the contributions made by our wonderful users. We'll use this section to call out some of these contributions! If you'd like to help out as well, please see the [CONTRIBUTING](CONTRIBUTING.md) guide. | |||
|
|||
- [@SalmanMohammadi](https:/salmanmohammadi) for adding a comprehensive end-to-end recipe for [PPO](recipes/lora_dpo_single_device.py) training to torchtune |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We need a better way to handle this. These are definitely some of the best contributions we have, but we should highlight everyone who contributes to the repo. What about axolotl's acknowledgements: https:/axolotl-ai-cloud/axolotl#contributing-?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't intend for this to be an exhaustive list, but rather a more high-signal place to list the most valuable contributions we've got. The axolotl approach works well for them, but in our case it'll be 50% various PyTorch devs which imo defeats the purpose. I agree it's not an approach we should continue indefinitely, but this gives direct access to a GitHub user profile which I think is more valuable (on the Axolotl page it's basically just a JPEG with a hyperlink to the contributors page). We can consider keeping only the ~5 latest community contributions here so it doesn't become overwhelming though
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Awesome, thanks for volunteering to keep this list updated at all times!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That came off snarkier than I intended
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Or did it...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It may also introduces some weirdness/politics in who decides what is important enough.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All contributors with the name Joe automatically go to the top of this list
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Anyone named Evan is permabanned
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Another idea would be to feature all contributors in the CONTRIBUTING.md file and just do a better job highlighting our added features on a monthly basis in the main README, with an emphasis on community contributions.
This would be like an "Updates" section on the README that holds a memory of 1-3 months or so.
Thanks to @SalmanMohammadi we now have a PPO recipe! We should show it off.
Also we should have added Qwen2 to our README already so doing that now too.
Also add acknowledgment on the readme to both @SalmanMohammadi for the PPO recipe and @fyabc for Qwen2 models.
(And fix a broken DPO hyperlink while I'm in here)