A high-throughput and memory-efficient inference and serving engine for LLMs
[CI] Fix attention benchmark smoke test (#45728)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com> Co-authored-by: Claude <noreply@anthropic.com>
Matthew Bonanni committed
Jun 16, 2026 at 09:43 UTC
040df8f2eadf164775d82a9972148070c8f48207
2 files changed
+2
-2
.buildkite/test-amd.yaml
+1
-1
@@ -2759,7 +2759,7 @@ steps:
2759
- vllm/_aiter_ops.py
2760
- vllm/platforms/rocm.py
2761
commands:
2762
- - python3 benchmarks/attention_benchmarks/benchmark.py --backends ROCM_ATTN ROCM_AITER_FA ROCM_AITER_UNIFIED_ATTN --batch-specs "8q1s1k" --repeats 1 --warmup-iters 1
2762
+ - python3 benchmarks/attention_benchmarks/benchmark.py --backends ROCM_ATTN ROCM_AITER_FA ROCM_AITER_UNIFIED_ATTN --batch-specs "8q1s1k"
2763
2764
#-------------------------------------------------------- mi355 · distributed --------------------------------------------------------#
2765
.buildkite/test_areas/benchmarks.yaml
+1
-1
@@ -23,4 +23,4 @@ steps:
23
- benchmarks/attention_benchmarks/
24
- vllm/v1/attention/
25
commands:
26
- - python3 benchmarks/attention_benchmarks/benchmark.py --backends flash flashinfer --batch-specs "8q1s1k" --repeats 1 --warmup-iters 1
26
+ - python3 benchmarks/attention_benchmarks/benchmark.py --backends flash flashinfer --batch-specs "8q1s1k"