A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 22:35 UTC
Revert "[compile][startup] Disable C++ compilation of symbolic shapes" (#22122)
Signed-off-by: Xiao Liu <xiszishu@gmail.com>
Xiao committed
Aug 2, 2025 at 09:03 UTC
554df8a6a2ed9007086f64768803ae4c780127bd
1 file changed
+2
-9
vllm/compilation/decorators.py
+2
-9
@@ -267,15 +267,8 @@ def _support_torch_compile(
267
code.co_filename)
268
return inline_call(parent, func, args, kwargs)
269
270
- # Disable the C++ compilation of symbolic shape guards. C++-fication
271
- # of symbolic shape guards can improve guard overhead. But, since
272
- # vllm skip guards anyways, setting this flag to False can improve
273
- # compile time.
274
- with torch._dynamo.config.patch("enable_cpp_symbolic_shape_guards",
275
- False), patch.object(
276
- InliningInstructionTranslator,
277
- 'inline_call',
278
- patched_inline_call):
270
+ with patch.object(InliningInstructionTranslator, 'inline_call',
271
+ patched_inline_call):
272
output = self.compiled_callable(*args, **kwargs)
273
return output
274