diff --git a/build_push.sh b/build_push.sh index 3d2f459..31c9bcd 100644 --- a/build_push.sh +++ b/build_push.sh @@ -45,6 +45,7 @@ services: - /root/nvidia-meeting/DeepSeek-V4-Pro-NVFP4:/model environment: - VLLM_USE_FLASHINFER_MOE_FP4=1 + - OMP_NUM_THREADS=128 deploy: resources: reservations: diff --git a/docker-compose.yml b/docker-compose.yml index c12f298..9d72f49 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -11,7 +11,7 @@ services: - -- environment: - HF_TOKEN=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 - - OMP_NUM_THREADS=64 + - OMP_NUM_THREADS=128 command: - /model - --trust-remote-code