[rollout] fix: vllm-rollout update weights

[Annarine]

What does this PR do?

Fix the bug, some model need to do some work after update weights;
Example: Ascend DeepSeekV3 mla : after update weights should tranpose

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, 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).)

[gemini-code-assist]

Code Review

This pull request adds a call to process_weights_after_loading after updating weights for non-FP8 models. This appears to be an important bugfix. However, the same logic is missing for the FP8 model code path, which could lead to incorrect model behavior. I've added a critical comment to highlight this and recommend applying the fix to the FP8 path as well to ensure correctness for all model types.

[gemini-code-assist]

The addition of process_weights_after_loading seems to be a necessary bugfix for correctly updating model weights. However, this logic is only applied to the non-FP8 code path. The FP8 path, handled in the if is_fp8_model(...) block (lines 263-267), also performs weight loading but misses this crucial post-processing step.

This omission could lead to incorrect behavior or bugs for FP8 models after weights are updated.

Please ensure process_weights_after_loading is also called for the FP8 model path. A possible solution is to duplicate these lines inside the if block, or refactor to call it once after the if-else statement.