EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Refactor tensor_parallel and comm_utils (#2814)

Browse Source 
### What this PR does / why we need it?
1. Move ops/comm_utils to ops/moe/comm_utils
2. Move distributed/tensor_parallel/gather_from_sequence_parallel_region
to ops/moe/comm_utils
3. Delete distributed/tensor_parallel

### Does this PR introduce _any_ user-facing change?
no

### How was this patch tested?
e2e & ut

- vLLM version: main
- vLLM main:
a1213fae5f

---------

Signed-off-by: wuweiqiang24 <1005334931@qq.com>
Signed-off-by: wuweiqiang24 <wuweiqiang11@huawei.com>
This commit is contained in:
wuweiqiang24
2025-09-11 21:26:36 +08:00
committed by GitHub
parent 0005479b9c
commit 9615dea3a7
6 changed files with 153 additions and 392 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/ut/ops/test_token_dispatcher.py
View File

@@ -348,7 +348,7 @@ class TestTokenDispatcherWithAll2AllV(TestBase):
self.mock_npu_moe_token_unpermute.return_value = torch.randn(8, 16)
# Mock async_all_to_all
patcher6 = patch('vllm_ascend.ops.comm_utils.async_all_to_all')
patcher6 = patch('vllm_ascend.ops.moe.comm_utils.async_all_to_all')
self.mock_async_all_to_all = patcher6.start()
self.addCleanup(patcher6.stop)
self.mock_async_all_to_all.return_value = (None, torch.randn(16, 16),