biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Attention] Refactor CUDA attention backend selection logic (#24794)

Browse Source 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
Matthew Bonanni
2025-11-11 06:40:44 -06:00
committed by GitHub
parent 2e78150d24
commit b30dfa03c5
61 changed files with 1338 additions and 1002 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/v1/spec_decode/test_mtp.py
View File

@@ -12,7 +12,7 @@ from tests.v1.attention.utils import (
create_standard_kv_cache_spec,
try_get_attention_backend,
)
from vllm.attention.backends.registry import _Backend
from vllm.attention.backends.registry import AttentionBackendEnum
from vllm.config import (
CacheConfig,
DeviceConfig,
@@ -177,7 +177,9 @@ def test_mtp_propose(num_speculative_tokens, monkeypatch):
sampling_metadata = mock.MagicMock()
# Setup attention metadata
attn_metadata_builder_cls, _ = try_get_attention_backend(_Backend.FLASH_ATTN)
attn_metadata_builder_cls, _ = try_get_attention_backend(
AttentionBackendEnum.FLASH_ATTN
)
attn_metadata_builder = attn_metadata_builder_cls(
kv_cache_spec=create_standard_kv_cache_spec(proposer.vllm_config),