biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[torch.compile] add logging for compilation time (#10941)

Browse Source 
Signed-off-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
youkaichao
2024-12-06 02:07:15 -08:00
committed by GitHub
parent db87eb6c67
commit b031a455a9
6 changed files with 75 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/config.py
View File

@@ -2281,6 +2281,7 @@ class CompilationConfig(BaseModel):
# keep track of enabled and disabled custom ops
enabled_custom_ops: Counter[str] = PrivateAttr
disabled_custom_ops: Counter[str] = PrivateAttr
compilation_time: float = PrivateAttr
# Per-model forward context
# Mainly used to store attention cls
@@ -2319,6 +2320,7 @@ class CompilationConfig(BaseModel):
self.enabled_custom_ops = Counter()
self.disabled_custom_ops = Counter()
self.static_forward_context = {}
self.compilation_time = 0.0
def init_backend(self) -> Union[str, Callable]:
if self.level == CompilationLevel.NO_COMPILATION: