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 14, 23:45 UTC
[CI] Parallelize Kernels MoE Test (#21764)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Jul 28, 2025 at 21:56 UTC
afa26075966301887a15f958a6aec0a89a3faacd
1 file changed
+4
-3
.buildkite/test-pipeline.yaml
+4
-3
@@ -403,17 +403,18 @@ steps:
403
- vllm/model_executor/layers/quantization
404
- tests/kernels/quantization
405
commands:
406
- - pytest -v -s kernels/quantization --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
406
+ - pytest -v -s kernels/quantization --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
407
parallelism: 2
408
409
-- label: Kernels MoE Test
409
+- label: Kernels MoE Test %N
410
mirror_hardwares: [amdexperimental]
411
source_file_dependencies:
412
- csrc/moe/
413
- tests/kernels/moe
414
- vllm/model_executor/layers/fused_moe/
415
commands:
416
- - pytest -v -s kernels/moe
416
+ - pytest -v -s kernels/moe --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
417
+ parallelism: 2
418
419
- label: Kernels Mamba Test
420
mirror_hardwares: [amdexperimental, amdproduction]