55 lines
1.3 KiB
YAML
55 lines
1.3 KiB
YAML
# compose.gpu.yml — Docker NVIDIA GPU overlay
|
|
#
|
|
# Adds NVIDIA GPU reservations to Peregrine services.
|
|
# Applied automatically by `make start PROFILE=single-gpu|dual-gpu` when Docker is detected.
|
|
# Manual: docker compose -f compose.yml -f compose.gpu.yml --profile single-gpu up -d
|
|
#
|
|
# Prerequisites:
|
|
# sudo nvidia-ctk runtime configure --runtime=docker
|
|
# sudo systemctl restart docker
|
|
#
|
|
services:
|
|
ollama:
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
device_ids: ["0"]
|
|
capabilities: [gpu]
|
|
|
|
ollama_research:
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
device_ids: ["1"]
|
|
capabilities: [gpu]
|
|
|
|
vision:
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
device_ids: ["0"]
|
|
capabilities: [gpu]
|
|
|
|
vllm:
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
device_ids: ["1"]
|
|
capabilities: [gpu]
|
|
|
|
finetune:
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
device_ids: ["0"]
|
|
capabilities: [gpu]
|