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 16, 23:15 UTC
[ci] Limit num gpus if specified for A100 (#5694)
Signed-off-by: kevin <kevin@anyscale.com>
Kevin H. Luu committed
Jun 19, 2024 at 16:30 UTC
949e49a6857080e36ecd62f6e193754290c7c43c
2 files changed
+2
-1
.buildkite/test-pipeline.yaml
+1
@@ -184,6 +184,7 @@ steps:
184
185
- label: Distributed Tests (A100)
186
gpu: a100
187
+ num_gpus: 4
188
commands:
189
# NOTE: don't test llama model here, it seems hf implementation is buggy
190
# see https://github.com/vllm-project/vllm/pull/5689 for details
.buildkite/test-template-aws.j2
+1
-1
@@ -75,7 +75,7 @@ steps:
75
- "'cd {{ (step.working_dir or default_working_dir) | safe }} && {{ step.command or (step.commands | join(' && ')) | safe }}'"
76
resources:
77
limits:
78
- nvidia.com/gpu: 8
78
+ nvidia.com/gpu: {{ step.num_gpus or 1 }}
79
volumeMounts:
80
- name: devshm
81
mountPath: /dev/shm