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, 01:00 UTC
[Misc] DeepEPHighThroughtput - Enable Inductor pass (#21311)
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com> Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Varun Sundar Rabindranath committed
Jul 22, 2025 at 12:05 UTC
8425f785ad58020ccda8b2d4d888f0a8be4af6c3
1 file changed
-3
vllm/platforms/cuda.py
-3
@@ -182,9 +182,6 @@ class CudaPlatformBase(Platform):
182
compilation_config.use_cudagraph = False
183
if model_config is not None:
184
model_config.enforce_eager = True
185
- # TODO (varun): Turning this ON gives incorrect results for the
186
- # Deepseek-V2-lite model.
187
- vllm_config.compilation_config.use_inductor = False
185
186
@classmethod
187
def get_current_memory_usage(cls,