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 12, 18:30 UTC
[docs] Fix cudagraph mode config (#29170)
Signed-off-by: angelayi <yiangela7@gmail.com>
Angela Yi committed
Nov 21, 2025 at 17:10 UTC
d5dbdbfcb2cfc2e4d82a1e2605576f1e4e440ca7
1 file changed
+1
-1
docs/design/debug_vllm_compile.md
+1
-1
@@ -9,7 +9,7 @@ TL;DR:
9
|----------|----------|-------------|
10
| --enforce-eager | enforce_eager=True | Turn off torch.compile and CUDAGraphs |
11
| -O.mode=0 | mode=CompilationMode.NONE | Turn off torch.compile only |
12
-| -O.cudagraph_mode=NONE | compilation_config=CompilationConfig(mode=CompilationMode.NONE) | Turn off CUDAGraphs only |
12
+| -O.cudagraph_mode=NONE | compilation_config=CompilationConfig(cudagraph_mode=CUDAGraphMode.NONE) | Turn off CUDAGraphs only |
13
| -O.backend=eager | compilation_config=CompilationConfig(backend='eager') | Turn off TorchInductor |
14
15
## vLLM-torch.compile overview