[Hardware][XPU] Register matmul and linear batch-invariant kernels for XPU (#49209)
Signed-off-by: tzielinski-habana <tomasz.zielinski@intel.com> Signed-off-by: Tomasz Zielinski <85164140+tzielinski-habana@users.noreply.github.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
co-authored by
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Claude Opus 5
Harry Mellor
parent
7894394b07
commit
ffe3bb3c72
@@ -2435,6 +2435,19 @@ class EngineArgs:
|
||||
self.attention_backend
|
||||
)
|
||||
|
||||
# Batch-invariant mode requires deterministic attention behavior.
|
||||
# If no backend is explicitly requested, prefer Triton Attention.
|
||||
if (
|
||||
envs.VLLM_BATCH_INVARIANT
|
||||
and attention_config.backend is None
|
||||
and current_platform.is_xpu()
|
||||
):
|
||||
attention_config.backend = AttentionBackendEnum.TRITON_ATTN
|
||||
logger.info(
|
||||
"VLLM_BATCH_INVARIANT is enabled and no attention backend was "
|
||||
"specified; defaulting to TRITON_ATTN."
|
||||
)
|
||||
|
||||
# TurboQuant requires FlashAttention 2 — FA3 boundary layers assert
|
||||
# FlashAttentionImpl which fails with TurboQuantAttentionImpl.
|
||||
if resolved_cache_dtype.startswith("turboquant_") and (
|
||||
|
||||
Reference in New Issue
Block a user