[Refactor]Refactor of vllm_ascend/distributed module

[luoxiaolin712]

What this PR does / why we need it?

Based on the RFC:#5604

This PR is a refactoring of vllm_ascend/distributed, moving all kv_transfer realtaed codes into a dedicated folder, which has already been done in vLLM

Does this PR introduce any user-facing change?

NA

How was this patch tested?

• vLLM version: v0.13.0
• vLLM main: vllm-project/vllm@2f4e654

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[gemini-code-assist]

Code Review

This pull request is a large-scale refactoring of the vllm_ascend/distributed module, moving files into a more organized directory structure and updating import paths accordingly. While this is a positive change for code organization, I've identified several incorrect import paths that will cause ImportError exceptions at runtime. These are critical issues that need to be fixed. I've provided specific comments and suggestions for each case.

[gemini-code-assist]

The import for _convert_to_bytes and _parse_global_segment_size is incorrect. They are being imported from the backend package, but the __init__.py for that package is empty, which will cause an ImportError. Please either update the __init__.py to expose these functions or change the import to be direct from their module.

Suggested change

	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend import ( # noqa: E402
	_convert_to_bytes, _parse_global_segment_size)
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.mooncake_backend import ( # noqa: E402
	_convert_to_bytes, _parse_global_segment_size)

[gemini-code-assist]

The import for Backend is incorrect. It is being imported from the backend package, but the __init__.py for that package is empty, which will cause an ImportError. Please either update the __init__.py to expose this class or change the import to be direct from its module.

Suggested change

	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend import Backend
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.backend import Backend

[gemini-code-assist]

The imports for Backend and MooncakeBackend are incorrect. They are being imported from the backend package, but the __init__.py for that package is empty, which will cause an ImportError. Please either update the __init__.py to expose these classes or change the imports to be direct from their respective modules.

Suggested change

	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend import Backend
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.memcache_backend import \
	MemcacheBackend
	from vllm_ascend.distributed.kvpool.backend.mooncake_backend import \
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend import \
	MooncakeBackend
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.backend import Backend
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.memcache_backend import \
	MemcacheBackend
	from vllm_ascend.distributed.kv_transfer.kv_pool.ascend_store.backend.mooncake_backend import \
	MooncakeBackend

[gemini-code-assist]

The import path for fc3_all_gather_and_maybe_unpad_impl is incorrect. This function was not moved to vllm_ascend.distributed.kv_transfer.utils.utils as part of the refactoring; it remains in vllm_ascend.distributed.utils. This change will cause an ImportError.

Suggested change

	from vllm_ascend.distributed.kv_transfer.utils.utils import fc3_all_gather_and_maybe_unpad_impl
	from vllm_ascend.distributed.utils import fc3_all_gather_and_maybe_unpad_impl

[luoxiaolin712]

The RFC has reached an agreement.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[wangxiyuan]

@LCAIZJ @liziyu179

[wangxiyuan]

please fix the merge conflict.

[lidenghui1110]

organized by function is ok to me, howerver, cpu_offload_manager in distributed directory is relied by cpu_offload_connector, maybe cpu_offload_manager should be moved into cpu_offload dir after this refactor?

[Pz1116]

organized by function is ok to me, howerver, cpu_offload_manager in distributed directory is relied by cpu_offload_connector, maybe cpu_offload_manager should be moved into cpu_offload dir after this refactor?

my bad, moving cpu_offload_manager under cpu_offload dir is exactly what we planed to do, but I missed this part when drawing the table in the RFC, we'll fix this shortly.

[LCAIZJ]

LGTM