32 lines
1.2 KiB
YAML
32 lines
1.2 KiB
YAML
networks:
|
|
traefik-exposed:
|
|
name: traefik-exposed
|
|
external: true
|
|
|
|
services:
|
|
qwen:
|
|
image: vllm/vllm-openai:latest
|
|
runtime: nvidia
|
|
restart: always
|
|
networks:
|
|
- traefik-exposed
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- capabilities: [ gpu ]
|
|
volumes:
|
|
- ~/.cache/huggingface:/root/.cache/huggingface
|
|
ports:
|
|
- "5432:5432"
|
|
command: --model Qwen/Qwen2.5-1.5B-Instruct --port 5432 --quantization bitsandbytes
|
|
labels:
|
|
- "traefik.enable=true"
|
|
- "traefik.http.routers.qwen.rule=Host(`odin.capgemini.com`) && PathPrefix(`/qwen`)"
|
|
- "traefik.http.routers.qwen.entrypoints=web"
|
|
- "traefik.http.services.qwen.loadbalancer.server.port=5432"
|
|
- "traefik.http.routers.qwen.middlewares=qwen-stripprefix"
|
|
- "traefik.http.middlewares.qwen-stripprefix.stripprefix.prefixes=/qwen"
|
|
# HTTPS is currently not supported on odin.
|
|
# - "traefik.http.routers.http-catchall.middlewares=redirect-to-https"
|
|
# - "traefik.http.middlewares.redirect-to-https.redirectscheme.scheme=https" |