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

[BugFix]Support redundant experts in EPLB (#3473)

Browse Source 
This PR adds support for redundant experts in the EPLB. 

Key points: 
- Use global_num_experts = num_experts + num_redundant_experts
consistently.
- Backward compatible when num_redundant_experts=0. 

Tested 
On a 16-rank setup (W8A8) with static EPLB and expert_map_path,
verifying router logits shape and successful requests.

- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

Signed-off-by: yechao237 <yechao20180411@gmail.com>
This commit is contained in:
yechao237
2025-10-18 00:09:16 +08:00
committed by GitHub
parent 07ca1b9b78
commit 4750d45d86
12 changed files with 23 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -257,7 +257,7 @@ class MockFusedMoEMethod(FusedMoEMethodBase):
class TestExpertsSelector:
@pytest.mark.parametrize("global_num_experts", [[256], [128]])
@pytest.mark.parametrize("global_num_experts", [256, 128])
def test_select_experts(self, mock_dist_env, mock_moe_env,
global_num_experts):