[megatron] feat: add script for qwen3_235b_grpo training on npu platform

[wangshuyang31]

What does this PR do?

Add concise overview of what this PR aims to achieve or accomplish. Reference related GitHub issues and PRs that help with the review.

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.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[gemini-code-assist]

Code Review

This pull request introduces a new training script for qwen3_235b_grpo on an NPU platform. The script configures various environment variables and hyperparameters for a verl.trainer.main_ppo training run. My review identified two critical issues in the shell script that would prevent it from executing correctly. One is a typo in a Hydra override argument, and the other is an incorrect placement of shell redirection, which would lead to some arguments being ignored. I have provided suggestions to resolve these issues.

[gemini-code-assist]

The shell redirection 2>&1 | tee ... is misplaced in the middle of the python3 command's arguments. This will cause all subsequent arguments (lines 154-155) to be piped to tee instead of being passed to the Python script. Additionally, the trailing backslash \ on the last line is a syntax error. The redirection should be moved to the end of the entire command to ensure all arguments are correctly passed to the script.

Suggested change

	trainer.device=npu 2>&1 | tee "logs/$(date +%Y%m%d_%H%M).log" \
	+actor_rollout_ref.rollout.engine_kwargs.vllm.compilation_config.cudagraph_capture_sizes="[8, 16, 32, 64, 128]" \
	+actor_rollout_ref.rollout.engine_kwargs.vllm.compilation_config.cudagraph_mode="FULL_DECODE_ONLY" \
	trainer.device=npu \
	+actor_rollout_ref.rollout.engine_kwargs.vllm.compilation_config.cudagraph_capture_sizes="[8, 16, 32, 64, 128]" \
	+actor_rollout_ref.rollout.engine_kwargs.vllm.compilation_config.cudagraph_mode="FULL_DECODE_ONLY" 2>&1 | tee "logs/$(date +%Y%m%d_%H%M).log"

[wucong25]

trainer.device=npu 2>&1 | tee "logs/$(date +%Y%m%d_%H%M).log" \这一行 2>&1 | tee "logs/$(date +%Y%m%d_%H%M).log应该放在结尾