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, 19:00 UTC
[CI Failure] Disable B200 tests while runner is broken (#31732)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Jan 5, 2026 at 11:50 UTC
eefa713a6605624c3cb2c538982bd37f3d56b187
1 file changed
+4
-1
.buildkite/test-pipeline.yaml
+4
-1
@@ -383,6 +383,7 @@ 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
387
source_file_dependencies:
388
- vllm/v1/attention
389
- tests/v1/attention
@@ -943,7 +944,7 @@ steps:
944
timeout_in_minutes: 30
945
working_dir: "/vllm-workspace/"
946
gpu: b200
946
- # optional: true
947
+ optional: true # TODO(mgoin): disable optional once runner is fixed
948
source_file_dependencies:
949
- csrc/quantization/fp4/
950
- csrc/attention/mla/
@@ -985,6 +986,7 @@ steps:
986
timeout_in_minutes: 40
987
working_dir: "/vllm-workspace/"
988
gpu: b200
989
+ optional: true # TODO(mgoin): disable optional once runner is fixed
990
source_file_dependencies:
991
- csrc/quantization/fp4/
992
- vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
@@ -1052,6 +1054,7 @@ steps:
1054
timeout_in_minutes: 60
1055
working_dir: "/vllm-workspace/"
1056
gpu: b200
1057
+ optional: true # TODO(mgoin): disable optional once runner is fixed
1058
source_file_dependencies:
1059
- tests/quantization/test_blackwell_moe.py
1060
- vllm/model_executor/models/deepseek_v2.py