[megatron] chore: clean legacy code path part 2, clean legacy CI

[ISEEKYAN]

What does this PR do?

this is one of a series PRs to clean the legacy megatron code path and make bridge the default path for megatron. #4496

This PR make clean CIs, including qwen2.5vl megatron, model merger, config converter

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 focuses on cleaning up legacy Megatron code paths and making mbridge the default, which is a positive step towards simplifying the codebase. The changes across documentation, shell scripts, and Python code are largely consistent with this goal. However, I've identified a critical issue in verl/models/mcore/registry.py where the logic for identifying Vision Language Models (VLMs) is flawed. This bug would cause VLMs to use an incorrect forward pass, breaking their functionality. I've provided suggestions to correct this. Please address these critical issues to ensure the new implementation works as expected for all model types.

[gemini-code-assist]

The check hf_config.architectures[0] in SupportedVLM is incorrect. SupportedVLM is an Enum class, and this check will always evaluate to False because hf_config.architectures[0] is a string, not an enum member. This will cause Vision Language Models (VLMs) to be incorrectly treated as standard language models, leading to the use of the wrong forward function.

To fix this, you should check against the values of the enum members.

Suggested change

	if hf_config.architectures[0] in SupportedVLM:
	if hf_config.architectures[0] in {item.value for item in SupportedVLM}:

[gemini-code-assist]

Similar to the previous comment, the check hf_config.architectures[0] in SupportedVLM is incorrect and will always evaluate to False. This will cause Vision Language Models (VLMs) to use the wrong fused forward function, as they will be misidentified as standard language models.

To fix this, you should check against the values of the enum members.

Suggested change

	if hf_config.architectures[0] in SupportedVLM:
	if hf_config.architectures[0] in {item.value for item in SupportedVLM}: