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 7, 22:45 UTC
[CI] Stabilizing ROCm amd-ci signal and minor name fix in upstream (#35008)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Feb 21, 2026 at 22:01 UTC
d403c1da1cd5c210581a2ed4c08c6b932b45186b
3 files changed
+4
-2
.buildkite/test-amd.yaml
+1
-1
@@ -67,7 +67,7 @@ steps:
67
timeout_in_minutes: 30
68
mirror_hardwares: [amdexperimental, amdproduction, amdtentative]
69
agent_pool: mi325_1
70
- grade: Blocking
70
+ # grade: Blocking
71
source_file_dependencies:
72
- vllm/
73
- tests/test_inputs.py
.buildkite/test_areas/distributed.yaml
+1
-1
@@ -209,7 +209,7 @@ steps:
209
- uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
210
- CROSS_LAYERS_BLOCKS=True bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
211
212
-- label: Pipeline + Context Parallelism (4 GPUs))
212
+- label: Pipeline + Context Parallelism (4 GPUs)
213
timeout_in_minutes: 60
214
working_dir: "/vllm-workspace/tests"
215
num_devices: 4
requirements/rocm-test.txt
+2
@@ -104,3 +104,5 @@ segmentation-models-pytorch==0.5.0
104
imagehash==4.3.2
105
# Required for bitsandbytes quantization test
106
bitsandbytes==0.49.2
107
+# Examples (tensorizer) tests
108
+tensorizer==2.10.1