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 15, 13:30 UTC
[easy] Pass compile_fx only the config patches (#16845)
Signed-off-by: rzou <zou3519@gmail.com>
Richard Zou committed
Apr 20, 2025 at 00:25 UTC
fe742aef5aaf406c62cafa248068818bfe517d6e
1 file changed
+1
-2
vllm/compilation/compiler_interface.py
+1
-2
@@ -167,8 +167,7 @@ class InductorAdaptor(CompilerInterface):
167
compiler_config: Dict[str, Any],
168
runtime_shape: Optional[int] = None
169
) -> Tuple[Optional[Callable], Optional[Any]]:
170
- from torch._inductor import config
171
- current_config = config.get_config_copy()
170
+ current_config = {}
171
from torch._inductor.compile_fx import compile_fx
172
173
# disable remote cache