[trainer] fix: Fix ClearML logging

[Tialo]

What does this PR do?

fix typo and use close because mark_completed closes the program

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, 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 correctly addresses two issues with ClearML logging: a typo in the logger key and the use of a deprecated method that could cause premature program termination. The fixes are sound. My review includes one high-severity recommendation to improve the robustness of the resource cleanup mechanism, which currently relies on the non-deterministic __del__ method.

[gemini-code-assist]

While this change correctly fixes the typo for the clearml logger, the overall approach of using __del__ for resource cleanup is unreliable and should be avoided.

The __del__ method is not guaranteed to be called in Python under all circumstances (e.g., with circular references, or during interpreter shutdown). Relying on it for critical cleanup tasks like finalizing loggers can lead to data loss or experiments being left in an inconsistent state.

A much safer and more idiomatic Python pattern is to provide an explicit finish() or close() method on the Tracking class and use it as a context manager. This ensures that cleanup is performed deterministically.

I strongly recommend refactoring this cleanup logic out of __del__. Here is an example of how this could be implemented:

class Tracking:
    # ... existing __init__ and log methods ...

    def finish(self):
        if "wandb" in self.logger:
            self.logger["wandb"].finish(exit_code=0)
        if "swanlab" in self.logger:
            self.logger["swanlab"].finish()
        if "vemlp_wandb" in self.logger:
            self.logger["vemlp_wandb"].finish(exit_code=0)
        if "tensorboard" in self.logger:
            self.logger["tensorboard"].finish()
        if "clearml" in self.logger:
            self.logger["clearml"].finish()
        if "trackio" in self.logger:
            self.logger["trackio"].finish()
        if "file" in self.logger:
            self.logger["file"].finish()

    def __enter__(self):
        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        self.finish()

    def __del__(self):
        # As a fallback, but ideally finish() is called explicitly
        # or via the context manager. Consider logging a warning here
        # if the object was not explicitly finished.
        self.finish()

This would require consumers of the Tracking class to use a with statement or call finish() explicitly, guaranteeing proper cleanup.