mirror of
https://github.com/furyhawk/home_stack.git
synced 2026-07-21 10:16:47 +00:00
29 lines
755 B
YAML
29 lines
755 B
YAML
services:
|
|
vllm:
|
|
image: docker.io/vllm/vllm-openai:cu130-nightly
|
|
command:
|
|
- --model
|
|
# - Qwen/Qwen3-4B-Instruct-2507
|
|
- Qwen/Qwen3-VL-2B-Instruct
|
|
- --gpu-memory-utilization
|
|
- "0.75"
|
|
- --max-model-len
|
|
- "2048"
|
|
ports:
|
|
- "9000:8000"
|
|
environment:
|
|
HF_TOKEN: ${HF_TOKEN}
|
|
volumes:
|
|
- ${HOME}/.cache/huggingface:/root/.cache/huggingface
|
|
ipc: host
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
# WARN: requires Docker Compose 2.24.2
|
|
# https://docs.docker.com/reference/compose-file/merge/#replace-value
|
|
devices: !override
|
|
- capabilities: ["gpu"]
|
|
driver: cdi
|
|
device_ids:
|
|
- nvidia.com/gpu=all
|