Skip to content

Conversation

@stefanwebb
Copy link
Contributor

@stefanwebb stefanwebb commented Nov 13, 2025

Oumi is an open source project for end-to-end foundation model development (data synthesis, fine-tuning, eval, quantization and distillation, etc.) with good traction (8.6k GitHub stars):

https://github.com/oumi-ai/oumi

This PR adds a link to our demo notebook showing how to do GRPO training with Oumi + OpenEnv + vLLM:

https://github.com/oumi-ai/oumi/blob/main/notebooks/Oumi%20-%20OpenEnv%20GRPO%20with%20trl.ipynb

Oumi is an open source project for end-to-end foundation model development (data synthesis, fine-tuning, eval, quantization and distillation, etc.) with good traction (8.6k GitHub stars):

https://github.com/oumi-ai/oumi

This PR adds a link to our demo notebook showing how to do GRPO training with Oumi + OpenEnv + vLLM
@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Nov 13, 2025
@stefanwebb stefanwebb changed the title Add Oumi to OpenEnv partner platforms -> README.md [README] Add link to Oumi notebook to OpenEnv partner platforms Nov 13, 2025
@burtenshaw burtenshaw self-requested a review November 13, 2025 19:35
@burtenshaw
Copy link
Collaborator

burtenshaw commented Nov 13, 2025

Really nice example. LGTM.

  • Out of curiosity, what does OUMI add to GRPO over vanilla TRL?
  • In the notebook, it might be helpful for users to show the reward curves, via a logging tool like trackio or just a plot image.

@stefanwebb
Copy link
Contributor Author

That's a great question! Oumi is working one higher level of abstraction over vanilla TRL. You can run TRL (or Verl) training via a declarative configuration and the CLI.

Also, Oumi is a library for the complete pipeline, so you could do some data synthesis, feed that to TRL + OpenEnv, run LLM-as-a-Judge, generate a revised prompt for data synthesis, etc. in the one system

@jspisak
Copy link
Contributor

jspisak commented Nov 14, 2025

love this @stefanwebb - thanks for doing this!

When you've finalized it let us know and one of us can merge.

@burtenshaw burtenshaw merged commit 2abede4 into meta-pytorch:main Nov 14, 2025
1 check passed
@stefanwebb
Copy link
Contributor Author

@burtenshaw thanks for the merge! :)

@stefanwebb stefanwebb deleted the patch-1 branch November 14, 2025 16:05
stefanwebb added a commit to stefanwebb/OpenEnv that referenced this pull request Nov 14, 2025
I see there is a second section in the README for tool integrations. Could we please add Oumi to this as well?

This relates to meta-pytorch#195 that added Oumi to the very top of the README
rycerzes pushed a commit to rycerzes/OpenEnv that referenced this pull request Nov 19, 2025
[README] Add link to Oumi notebook to OpenEnv partner platforms
rycerzes pushed a commit to rycerzes/OpenEnv that referenced this pull request Nov 19, 2025
I see there is a second section in the README for tool integrations. Could we please add Oumi to this as well?

This relates to meta-pytorch#195 that added Oumi to the very top of the README
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants