xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-19 00:17:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

Log how much time loading a compiled artifact takes (#16848)

Browse Source 
Signed-off-by: rzou <zou3519@gmail.com>
This commit is contained in:
Richard Zou 2025-04-19 12:50:46 -04:00 committed by GitHub
parent d6195a748b
commit 682e0b6d2f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/compilation/backends.py
View File

@ -110,10 +110,14 @@ class CompilerManager:
compiled_graph = self.load(graph, example_inputs, graph_index, compiled_graph = self.load(graph, example_inputs, graph_index,
runtime_shape) runtime_shape)
if compiled_graph is not None: if compiled_graph is not None:
if graph_index == 0: if graph_index == num_graphs - 1:
# adds some info logging for the first graph # after loading the last graph for this shape, record the time.
logger.info("Directly load the compiled graph for shape %s " # there can be multiple graphs due to piecewise compilation.
"from the cache", str(runtime_shape)) # noqa now = time.time()
elapsed = now - compilation_start_time
logger.info(
"Directly load the compiled graph(s) for shape %s "
"from the cache, took %.3f s", str(runtime_shape), elapsed)
return compiled_graph return compiled_graph
# no compiler cached the graph, or the cache is disabled, # no compiler cached the graph, or the cache is disabled,