[Misc] Clean up more utils

[DarkLight1337]

Purpose

Part of #26900

• vllm.utils.init_cached_hf_modules -> vllm.utils.import_utils.init_cached_hf_modules
• vllm.utils.import_pynvml -> vllm.utils.import_utils.import_pynvml
• vllm.utils.get_mp_context -> vllm.utils.system_utils.get_mp_context
• vllm.utils.kill_process_tree -> vllm.utils.system_utils.kill_process_tree
• vllm.utils.set_ulimit -> vllm.utils.system_utils.set_ulimit
• vllm.utils.run_method -> vllm.v1.serial_utils.run_method
• vllm.utils.check_use_alibi -> ModelConfig.uses_alibi
• vllm.utils.enable_trace_function_call_for_thread -> VllmConfig.enable_trace_function_call_for_thread
• vllm.utils.argparse_utils.StoreBoolean is no longer used so I removed it

Test Plan

Test Result

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request focuses on cleaning up and reorganizing utility functions within the vLLM project. It involves moving functions from vllm.utils to more specific submodules like vllm.utils.import_utils and vllm.utils.system_utils, removing unused code, and making minor adjustments to improve code clarity and maintainability. I have identified a critical issue where a function is being moved without updating its call site within the same file. This needs to be addressed to ensure the code functions correctly after the changes.

[chatgpt-codex-connector]

💡 Codex Review

vllm/vllm/v1/worker/worker_base.py

Lines 187 to 194 in 88c0da5

	if vllm_config.model_config is not None:
	# it can be None in tests
	trust_remote_code = vllm_config.model_config.trust_remote_code
	if trust_remote_code:
	# note: lazy import to avoid importing torch before initializing
	from vllm.utils import init_cached_hf_modules
	init_cached_hf_modules()

Update init_cached_hf_modules import in worker wrapper

When trust_remote_code is enabled, the worker wrapper lazily executes from vllm.utils import init_cached_hf_modules. This helper was moved into vllm.utils.import_utils and is no longer exported from vllm.utils, so the import now raises ImportError and remote-code initialization fails. The lazy import should target the new module that defines init_cached_hf_modules.

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[hmellor]

Strange that organising the utils would cause this to be required, is the mocking not working properly for this one?

[DarkLight1337]

Model executor (imported by docs) now imports serial_utils which in turn imports cloudpickle.

[hmellor]

Ok, these are relatively cheap to install, so it's not a big problem.

It'd be nice to figure out how to remove them in future.