[hardware] refactor: replace device_name with config.trainer.device

[sunyi0505]

What does this PR do?

In some methods, the get_device() method is redundant, and we plan to replace get_deivce with config.trainer.device

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
  • 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.

[gemini-code-assist]

Code Review

This pull request refactors the code to replace device_name with config.trainer.device. The changes improve code maintainability by centralizing device configuration. However, the introduction of hardcoded values for the device in multiple files reduces the flexibility of the code. I have added comments to address these issues.

[gemini-code-assist]

The default value for device_name is hardcoded to cuda. If the intention is to support other devices like npu, this should be configurable or dynamically determined based on the available hardware. Otherwise, it will cause the program to crash if cuda is not available.

Suggested change

	self.device_name = kwargs.get("device_name", "cuda")
	self.device_name = kwargs.get("device_name", self.config.trainer.device)

[gemini-code-assist]

Hardcoding the device_name in wg_kwargs restricts the flexibility of the code. It's better to pass the device configuration through the config object to allow users to easily switch between different devices without modifying the code.

[gemini-code-assist]

Adding trainer.device=npu directly to the script makes the script specific to NPU devices. It would be more flexible to allow the user to specify the device via an environment variable or command-line argument. This way, the same script can be used for different devices without modification.

[eric-haibin-lin]

cc @hongpeng-guo

[hongpeng-guo]

@as12138 I agree it's a good refactor to move device_name to the kwargs from a positional args. Could you please add this API change to this issue: #2528, so that any custom extensions of single controller can have a source of API changes upon next verl version.

[sunyi0505]

Of course, I have already added it.