Files
nvfp4-megamoe-kernel/docker-compose.yml

39 lines
1.0 KiB
YAML

services:
vllm:
build:
context: .
dockerfile: Dockerfile
ports:
- "8000:8000"
environment:
- OMP_NUM_THREADS=128
- CUDA_LAUNCH_BLOCKING=0
- PYTHONUNBUFFERED=1
- VLLM_RPC_TIMEOUT_MS=600000
- CLAWMINE_DEBUG=1
# Don't set VLLM_NVFP4_GEMM_BACKEND - our CuTeDSL kernel auto-selects on Blackwell
command:
- /model
- --trust-remote-code
- --enable-expert-parallel
- --tensor-parallel-size=8
- --compilation-config={"cudagraph_mode":"NONE","custom_ops":["all"]}
- --tokenizer-mode=deepseek_v4
- --tool-call-parser=deepseek_v4
- --enable-auto-tool-choice
- --reasoning-parser=deepseek_v4
- --moe-backend=cutedsl
- --gpu-memory-utilization=0.9
- --max-model-len=512
- --host=0.0.0.0
- --port=8000
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
volumes:
- /root/nvidia-meeting/DeepSeek-V4-Pro-NVFP4:/model:ro