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

[Cleanup] Remove unused attn_metadata parameter from Proposer classes (#4862)

Browse Source 
The `attn_metadata` is not used by any draft proposer, so we can remove
it.


- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: Jade Zheng <zheng.shoujian@outlook.com>
This commit is contained in:
Jade Zheng
2025-12-15 21:21:38 +08:00
committed by GitHub
parent a9625851ef
commit c064d11fd7
7 changed files with 1 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/ut/spec_decode/test_mtp_proposer.py
View File

@@ -238,7 +238,6 @@ class TestMtpProposer:
proposer.speculative_config = MagicMock(
disable_padded_drafter_batch=False)
proposer.pcp_size = mock_deps.runner.pcp_size
proposer._get_attn_metadata = MagicMock(return_value=MagicMock())
proposer.prepare_next_token_ids_padded = MagicMock(
return_value=(torch.tensor([101, 200, 302]), 3))
proposer.prepare_inputs_padded = MagicMock(
@@ -261,7 +260,6 @@ class TestMtpProposer:
proposer.prepare_next_token_ids_padded.assert_called_once()
proposer.prepare_inputs_padded.assert_called_once()
proposer._get_attn_metadata.assert_called_once()
proposer._propose.assert_called_once()
assert torch.equal(draft_token_ids, proposer._propose.return_value)