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

Revert "[Refactor][EAGLE] 8/N delete mtp_proposer" (#7030)

Browse Source 
Reverts vllm-project/vllm-ascend#7016
It breaks E2E test
- vLLM version: v0.16.0
- vLLM main:
4034c3d32e
This commit is contained in:
wangxiyuan
2026-03-06 11:24:05 +08:00
committed by GitHub
parent 8c2c82f3e1
commit 16c3b0b822
6 changed files with 931 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm_ascend/attention/attention_v1.py
View File

@@ -324,11 +324,7 @@ class AscendAttentionMetadataBuilder(AttentionMetadataBuilder[AscendMetadata]):
common_attn_metadata: AscendCommonAttentionMetadata,
attn_state: AscendAttentionState = AscendAttentionState.DecodeOnly,
):
if attn_state in (
AscendAttentionState.DecodeOnly,
AscendAttentionState.ChunkedPrefill,
AscendAttentionState.SpecDecoding,
):
if attn_state in (AscendAttentionState.DecodeOnly, AscendAttentionState.ChunkedPrefill):
attn_metadata = self.build(
common_prefix_len=0,
common_attn_metadata=common_attn_metadata,