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

[torch.compile] Fix tests for torch==2.9 inductor partition (#26116)

Browse Source 
Signed-off-by: ProExpertProg <lgovedic@redhat.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
This commit is contained in:
Luka Govedič
2025-10-14 19:55:02 -04:00
committed by GitHub
parent 579d2e5458
commit 2dcd12d357
8 changed files with 138 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/config/compilation.py
View File

@@ -201,7 +201,7 @@ class CompilationConfig:
(it sees a part of the graph). The backend can not be custom for compilation
level 3, i.e. the backend must be either eager or inductor. Furthermore,
compilation is only piecewise if splitting ops is set accordingly and
use_inductor_cudagraphs_partition is off. Note that the default options for
use_inductor_graph_partition is off. Note that the default options for
splitting ops are sufficient for piecewise compilation.
"""
custom_ops: list[str] = field(default_factory=list)
@@ -431,6 +431,7 @@ class CompilationConfig:
factors.append(self.custom_ops)
factors.append(self.splitting_ops)
factors.append(self.use_inductor)
factors.append(self.use_inductor_graph_partition)
factors.append(self.inductor_compile_config)
factors.append(self.inductor_passes)
factors.append(self.pass_config.uuid())