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 17, 08:00 UTC
Fix bad assert in initialize_cluster if PG already exists (#526)
Antoni Baum committed
Jul 19, 2023 at 23:17 UTC
c487a221ee27141170592d440cc973905a42fd97
1 file changed
+5
-3
vllm/engine/ray_utils.py
+5
-3
@@ -81,9 +81,11 @@ def initialize_cluster(
81
# Verify that we can use the placement group.
82
gpu_bundles = 0
83
for bundle in bundles:
84
- assert bundle.get("GPU", 0) > 1, (
85
- "Placement group bundles cannot have more than 1 GPU")
86
- if bundle.get("GPU", 0):
84
+ bundle_gpus = bundle.get("GPU", 0)
85
+ if bundle_gpus > 1:
86
+ raise ValueError(
87
+ "Placement group bundle cannot have more than 1 GPU.")
88
+ if bundle_gpus:
89
gpu_bundles += 1
90
if parallel_config.world_size > gpu_bundles:
91
raise ValueError(