iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[UX] Bump version in CG memory profiling log message (#40465)

Browse Source 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni
2026-04-21 15:26:06 +00:00
committed by GitHub
parent f819265a4a
commit ab5666eb7c
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/v1/worker/gpu_worker.py
+2 -2
View File
@@ -456,7 +456,7 @@ class Worker(WorkerBase):
logger.info(
"CUDA graph memory profiling is enabled "
"(VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=1). "
"This will become the default in v0.19. "
"This will become the default in v0.21. "
"The current --gpu-memory-utilization=%.4f is equivalent "
"to --gpu-memory-utilization=%.4f without CUDA graph "
"memory profiling. To maintain the same effective KV "
@@ -472,7 +472,7 @@ class Worker(WorkerBase):
1.0,
)
logger.info(
"In v0.19, CUDA graph memory profiling will be enabled "
"In v0.21, CUDA graph memory profiling will be enabled "
"by default (VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=1), "
"which more accurately accounts for CUDA graph memory "
"during KV cache allocation. To try it now, set "