[docker] Update stable image to vllm==0.15.1

[Begunner]

What does this PR do?

Update stable image to vllm==0.15.1
Modify necessary workflows

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, veomni, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data, cfg, reward
  • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.

API and Usage Example

Demonstrate how the API changes if any, and provide usage example(s) if possible.

# Add code snippet or script demonstrating how to use this

Design & Code Changes

Demonstrate the high-level design if this PR is complex, and list the specific changes.

Checklist Before Submitting

Important

Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: ...
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace. (If not accessible, please try the Feishu group (飞书群).)
• If your PR is related to the recipe submodule, please also update the reference to the submodule commit via git submodule update --remote or cd recipe && git pull origin main.

[gemini-code-assist]

Code Review

This pull request updates the stable Docker images to use vllm==0.15.1, simplifying the installation process by using pip instead of building from source. The changes also include pinning the transformers library version for better reproducibility and updating tests to accommodate the new dependencies. My main concern is a potential issue in docker/Dockerfile.stable.vllm where installing vllm might override the specified torch version, leading to a build failure in subsequent steps. I've provided a suggestion to address this.

[gemini-code-assist]

Installing vllm without the --no-deps flag may lead to its dependencies, including torch, being installed. This could potentially override the specific torch==2.9.1 version installed in the preceding step. If the torch version changes, the sed command on line 35 will fail as its path is hardcoded for torch-2.9.1+cu129.dist-info. The previous installation method for vllm used --no-deps, suggesting that dependencies are managed elsewhere. To prevent the build from breaking, you should add the --no-deps flag to this installation command.

RUN pip install --no-deps vllm==0.15.1

[thvasilo]

+1 here, it's usually safer to fully determine vllm 0.15.1 dependencies, install those first, and install vllm with --no-deps.

You can figure out vllm deps using something like

uv venv --python 3.12 --seed
source .venv/bin/activate
uv pip install vllm==0.15.1
uv pip show vllm

and can then pin the specific version shown by uv pip list

[thvasilo]

Thanks for the update, as confirmation I've tested vllm 0.15.0 myself and am able to run jobs with current dev branch of verl.

[thvasilo]

I think this can be removed as it not all users need vim and adds weight to iamge?