Skip to content

[Doc] Update README and fix a few installation related issues#347

Merged
listar2000 merged 3 commits into
rllm-org:mainfrom
listar2000:fix-readme-with-install-guide
Dec 16, 2025
Merged

[Doc] Update README and fix a few installation related issues#347
listar2000 merged 3 commits into
rllm-org:mainfrom
listar2000:fix-readme-with-install-guide

Conversation

@listar2000
Copy link
Copy Markdown
Collaborator

What does this PR do?

@listar2000 listar2000 marked this pull request as ready for review December 15, 2025 05:01
Comment thread README.md Outdated
## Getting Started 🎯

### Step 1: Building rLLM
rLLM requires `Python >= 3.11`. You can install it either directly via pip or build from source.
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

3.10 is fine, only tinker requires 3.11.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we can just enforce the user to use 3.11 now, given that Tinker has this requirement. This make sure user don't have to switch when they want to switch engine.

Comment thread README.md

```bash
uv pip install "git+https://github.com/rllm-org/rllm.git"
uv pip install "rllm[verl] @ git+https://github.com/rllm-org/rllm.git"
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

not sure if there is any use case for a direct installation @jeffreysijuntan
and if we do we should change DatasetRegistry to cache the datasets and register file to some user location like ~./rllm

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Currently no, but I think it is a good idea to cache datasets to ~/.rllm anyway so we should support this feature.

Comment thread README.md
**Step 2: Install rLLM with Training Backend**

rLLM supports two training backends: `verl` and `Tinker`. Choose one based on your needs.
rLLM supports two training backends: `verl` and `tinker`. Choose one based on your needs.
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

they can always install both e.g., uv pip install -e .[verl,tinker]

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah but I assume in most cases ppl would stick to only one of them maybe?

Comment thread README.md Outdated

# Install verl
bash scripts/install_verl.sh
# can add --torch-backend=auto to automatically detect the CUDA backend
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

in my testing, 'auto' did not work, ideally the user would always set this e.g., even in the cpu case they can set --torch-backend=cpu

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

let's remove this comment then.

@listar2000
Copy link
Copy Markdown
Collaborator Author

Thx for the review and great suggestions @kylemontgomery1. Since this PR relates to the crucial README, I would wait and see if @jeffreysijuntan has any comments (& response to some of the points you've raised). Will add a followup commit tmr.

@listar2000 listar2000 merged commit b774a82 into rllm-org:main Dec 16, 2025
1 check passed
@listar2000 listar2000 deleted the fix-readme-with-install-guide branch December 16, 2025 21:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants