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

[V0 deprecation] Remove _VLLM_V1 suffixes from attention backend names (#25489)

Browse Source 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
This commit is contained in:
Matthew Bonanni
2025-09-25 13:37:50 -04:00
committed by GitHub
parent 71b25b0d48
commit 3468f17ebe
42 changed files with 131 additions and 174 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/platforms/rocm.py
View File

@@ -218,8 +218,7 @@ class RocmPlatform(Platform):
raise ValueError(
f" The selected backend, {selected_backend.name},"
f"does not support block size {block_size}.")
if selected_backend in (_Backend.ROCM_AITER_MLA,
_Backend.ROCM_AITER_MLA_VLLM_V1):
if selected_backend == _Backend.ROCM_AITER_MLA:
if block_size == 1:
logger.info("Using AITER MLA backend on V1 engine.")
return "vllm.v1.attention.backends.mla.rocm_aiter_mla.AiterMLABackend" # noqa: E501
@@ -240,7 +239,7 @@ class RocmPlatform(Platform):
elif (envs.VLLM_ROCM_USE_AITER and
envs.VLLM_USE_AITER_UNIFIED_ATTENTION) or \
envs.VLLM_V1_USE_PREFILL_DECODE_ATTENTION or \
selected_backend == _Backend.ROCM_ATTN_VLLM_V1:
selected_backend == _Backend.ROCM_ATTN:
# rocm specific backend, with aiter and/or
# triton prefix-prefill
logger.info("Using Rocm/Aiter Attention backend on V1 engine.")