xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-17 04:55:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

[torch.compile] fix shape specialization (#10722)

Browse Source 
Signed-off-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
youkaichao 2024-11-27 10:16:10 -08:00 committed by GitHub
parent 308cc5e21e
commit c411def234
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/config.py
View File

@ -2151,7 +2151,7 @@ class CompilationConfig(BaseModel):
use_inductor: bool = True use_inductor: bool = True
inductor_specialize_for_cudagraph_no_more_than: Optional[int] = None inductor_specialize_for_cudagraph_no_more_than: Optional[int] = None
inductor_compile_sizes: Optional[List[int]] = Field(default_factory=dict) inductor_compile_sizes: Optional[List[int]] = Field(default=None)
inductor_compile_config: Dict = Field(default_factory=dict) inductor_compile_config: Dict = Field(default_factory=dict)
inductor_passes: Dict[str, str] = Field(default_factory=dict) inductor_passes: Dict[str, str] = Field(default_factory=dict)
@ -2290,9 +2290,8 @@ class CompilationConfig(BaseModel):
if x <= self.inductor_specialize_for_cudagraph_no_more_than if x <= self.inductor_specialize_for_cudagraph_no_more_than
] ]
else: else:
assert self.inductor_compile_sizes is not None, ( if self.inductor_compile_sizes is None:
"inductor_compile_sizes should not be None when " self.inductor_compile_sizes = []
"inductor_specialize_for_cudagraph_no_more_than is None")
self.compile_sizes = self.inductor_compile_sizes self.compile_sizes = self.inductor_compile_sizes