Davide Polonio
ebc71492c3
- Set HIP_VISIBLE_DEVICES=0 to use only the discrete GPU (gfx1201). llama.cpp was trying to split layers across the iGPU (gfx1036) which caused segfaults when loading the multimodal projector. - Restore --mmproj for both HF models (multimodal works correctly with single GPU). - Keep qwen3.5:9b disabled (Ollama-extracted GGUF uses old mrope_sections key format incompatible with this llama.cpp build).
91 lines
2.1 KiB
YAML
91 lines
2.1 KiB
YAML
services:
|
|
app:
|
|
build: .
|
|
image: local/llama-swap-rocm:latest
|
|
restart: unless-stopped
|
|
hostname: ollama
|
|
container_name: ollama
|
|
volumes:
|
|
- "/srv/docker/ollama/data/models:/models:ro"
|
|
- "./llama-swap.yaml:/etc/llama-swap/config.yaml:ro"
|
|
environment:
|
|
- HIP_VISIBLE_DEVICES=0
|
|
devices:
|
|
- "/dev/kfd:/dev/kfd"
|
|
- "/dev/dri:/dev/dri"
|
|
group_add:
|
|
- video
|
|
- render
|
|
networks:
|
|
internal:
|
|
ipv4_address: 172.23.0.5
|
|
logging:
|
|
driver: "json-file"
|
|
options:
|
|
mode: "non-blocking"
|
|
max-size: "10m"
|
|
max-file: "3"
|
|
|
|
tailscale:
|
|
hostname: ollama
|
|
image: tailscale/tailscale:latest
|
|
restart: unless-stopped
|
|
healthcheck:
|
|
test: ["CMD-SHELL", "tailscale status"]
|
|
interval: 1s
|
|
timeout: 5s
|
|
retries: 60
|
|
volumes:
|
|
- /srv/docker/ollama/tailscale:/var/lib
|
|
- /lib/modules:/lib/modules:ro
|
|
devices:
|
|
- /dev/net/tun:/dev/net/tun
|
|
cap_add:
|
|
- net_admin
|
|
- sys_module
|
|
- net_raw
|
|
command: tailscaled
|
|
networks:
|
|
- internal
|
|
logging:
|
|
driver: "json-file"
|
|
options:
|
|
mode: "non-blocking"
|
|
max-size: "10m"
|
|
max-file: "3"
|
|
|
|
reverse_proxy:
|
|
image: caddybuilds/caddy-namecheap:2-alpine
|
|
restart: unless-stopped
|
|
network_mode: service:tailscale
|
|
volumes:
|
|
- ./Caddyfile:/etc/caddy/Caddyfile:ro
|
|
- /srv/docker/ollama/caddy/config/:/config/caddy:rw
|
|
- /srv/docker/ollama/caddy/data/:/data/caddy:rw
|
|
- /srv/docker/ollama/caddy/share/:/usr/share/caddy:rw
|
|
env_file:
|
|
- caddy.env
|
|
healthcheck:
|
|
test: ["CMD", "wget", "--spider", "-q", "https://ollama.lan.poldebra.me"]
|
|
interval: 10s
|
|
timeout: 30s
|
|
retries: 5
|
|
start_period: 90s
|
|
depends_on:
|
|
app:
|
|
condition: service_started
|
|
tailscale:
|
|
condition: service_healthy
|
|
logging:
|
|
driver: "json-file"
|
|
options:
|
|
mode: "non-blocking"
|
|
max-size: "10m"
|
|
max-file: "3"
|
|
|
|
networks:
|
|
internal:
|
|
ipam:
|
|
config:
|
|
- subnet: 172.23.0.0/24
|