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, 03:15 UTC
[V1] log GPU blocks num for MultiprocExecutor (#11656)
WangErXiao committed
Jan 4, 2025 at 08:13 UTC
e5d7ed0c5374d38e75a8ef0243cc348f0f6f9185
1 file changed
+1
vllm/v1/executor/multiproc_executor.py
+1
@@ -95,6 +95,7 @@ class MultiprocExecutor(Executor):
95
Initialize the KV caches and begin the model execution loop of the
96
underlying workers.
97
"""
98
+ logger.info("# GPU blocks: %d", num_gpu_blocks)
99
self.collective_rpc("initialize_cache", args=(num_gpu_blocks, ))
100
self.collective_rpc("compile_or_warm_up_model")
101