24 lines
673 B
YAML
24 lines
673 B
YAML
services:
|
|
vllm:
|
|
image: vllm/vllm-openai:latest
|
|
container_name: vllm
|
|
network_mode: bridge
|
|
environment:
|
|
- HF_TOKEN=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
|
|
# - PYTORCH_ALLOC_CONF=expandable_segments:True
|
|
# - VLLM_DEVICE=cpu
|
|
ipc: host
|
|
command: --model Qwen/Qwen2.5-14B-Instruct
|
|
ports:
|
|
- "8000:8000"
|
|
volumes:
|
|
- /Var/lib/docker/volumes/vllm/huggingface:/root/.cache/huggingface
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
count: all
|
|
capabilities: [ gpu ]
|
|
runtime: nvidia
|
|
# command: --model NVFP4/Qwen3-Coder-30B-A3B-Instruct-FP4 |