[ci, training_utils] fix: get_nccl_backend default to nccl

[HollowMan6]

What does this PR do?

Fix CI error https://github.com/volcengine/verl/actions/runs/19335994615/job/55311051696

Note that in e2e_fully_async_policy CI environment, Gloo is not available, https://github.com/volcengine/verl/actions/runs/19345818399/job/55345837590 so we simply default by nccl to avoid any possible error here.

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. (If not accessible, please try the Feishu group (飞书群).)

[gemini-code-assist]

Code Review

This pull request aims to fix a CI error by enabling the gloo backend for CPU parameter synchronization in the fully asynchronous policy. The change correctly adds cpu:gloo to the backend configuration for ray.util.collective.create_collective_group. However, the implementation has a critical flaw: it will cause a RuntimeError on CPU-only environments because it unconditionally attempts to get an NCCL/HCCL backend. My review includes suggestions to fix this by conditionally constructing the backend string, ensuring the code works correctly on all hardware configurations.

[gemini-code-assist]

Code Review

The pull request correctly addresses a crash on CPU-only environments by defaulting to the 'gloo' backend. However, this change introduces a naming inconsistency, as the function get_nccl_backend now returns backends other than NCCL. I've added a high-severity comment recommending a rename of the function in a follow-up change to maintain code clarity.

[gemini-code-assist]

Code Review

This pull request refactors the get_nccl_backend function to get_distributed_backend, making the name more generic and improving robustness by defaulting to the 'gloo' backend for CPU-only environments instead of raising an error. The changes are applied consistently across the codebase. I have one piece of feedback regarding an outdated docstring that should be updated to align with these changes.

[gemini-code-assist]

Code Review

This pull request correctly addresses a CI failure by introducing get_distributed_backend to handle distributed training on CPU-only machines, defaulting to the gloo backend. The refactoring from get_nccl_backend is a good improvement. My review focuses on a few instances where the construction of the backend string for init_process_group could be more robust. Specifically, in a CPU-only environment, the current logic produces a redundant backend string. I've provided suggestions to handle this case more cleanly.

[HollowMan6]

/gemini review

[gemini-code-assist]

Code Review

This pull request aims to fix a CI error by changing how the distributed backend is selected. However, the change introduces a critical bug by defaulting to the 'nccl' backend, which will cause failures on CPU-only systems. My review provides a code suggestion to correct this by restoring explicit hardware checks, ensuring the code works correctly across all environments.