@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 10b67d865d
Fork 0

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 16, 08:15 UTC

[Bugfix] SymIntArrayRef expected to contain concrete integers (#10170)

Signed-off-by: Bill Nell <bill@neuralmagic.com>

bnellnm committed Nov 8, 2024 at 17:44 UTC 10b67d865d92e376956345becafc249d4c3c0ab7
1 file changed +2 -1
vllm/compilation/backends.py
+2 -1
@@ -329,7 +329,8 @@ class PiecewiseCompileInterpreter(torch.fx.Interpreter):
329 self.fake_mode.from_tensor(t) if isinstance(t, torch.Tensor) else t
330 for t in args
331 ]
332 - return super().run(*fake_args)
332 + with self.fake_mode:
333 + return super().run(*fake_args)
334
335 def call_module(self, target: torch.fx.node.Target,
336 args: Tuple[torch.fx.node.Argument,