A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 2, 17:01 UTC
[Misc][Rocm] Remove redundant `AiterUnifiedAttentionBackend` block size log (#43664)
Signed-off-by: NickLucche <nlucches@redhat.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Nicolò Lucchesi committed
May 28, 2026 at 05:19 UTC
413ac5c0701a86f17f56cd8c242894ed443d14ce
1 file changed
-3
vllm/v1/attention/backends/rocm_aiter_unified_attn.py
-3
@@ -30,9 +30,6 @@ class RocmAiterUnifiedAttentionBackend(RocmAttentionBackend):
30
31
@classmethod
32
def get_preferred_block_size(cls, default_block_size: int) -> int:
33
- logger.warning_once(
34
- "[ROCM_AITER_UNIFIED_ATTN]: Setting kv cache block size to 64."
35
- )
33
return 64
34
35
@classmethod