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 10, 17:45 UTC
Revert "[CI Failure] Disable B200 tests while runner is broken" (#31750)
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Michael Goin committed
Jan 5, 2026 at 20:26 UTC
ccb309a9640f90ca288ecea817fff7cc9474dbf5
1 file changed
-4
.buildkite/test-pipeline.yaml
-4
@@ -383,7 +383,6 @@ steps:
383
- label: V1 Test attention (B200) # 10min
384
timeout_in_minutes: 30
385
gpu: b200
386
- optional: true # TODO(mgoin): disable optional once runner is fixed
386
source_file_dependencies:
387
- vllm/v1/attention
388
- tests/v1/attention
@@ -944,7 +943,6 @@ steps:
943
timeout_in_minutes: 30
944
working_dir: "/vllm-workspace/"
945
gpu: b200
947
- optional: true # TODO(mgoin): disable optional once runner is fixed
946
source_file_dependencies:
947
- csrc/quantization/fp4/
948
- csrc/attention/mla/
@@ -986,7 +984,6 @@ steps:
984
timeout_in_minutes: 40
985
working_dir: "/vllm-workspace/"
986
gpu: b200
989
- optional: true # TODO(mgoin): disable optional once runner is fixed
987
source_file_dependencies:
988
- csrc/quantization/fp4/
989
- vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
@@ -1054,7 +1051,6 @@ steps:
1051
timeout_in_minutes: 60
1052
working_dir: "/vllm-workspace/"
1053
gpu: b200
1057
- optional: true # TODO(mgoin): disable optional once runner is fixed
1054
source_file_dependencies:
1055
- tests/quantization/test_blackwell_moe.py
1056
- vllm/model_executor/models/deepseek_v2.py