Spaces:
Sleeping
Sleeping
SpinalCord LLM: training, dashboard, speculative decoding, deploy docs, early-exit brain (PyTorch)
f52586c | # SpinalCord test UI behind nginx, API proxied to llama-server on the host. | |
| # | |
| # 1) On the host, start llama-server (e.g. dashboard/run_dashboard_prod.bat) on port 8080. | |
| # 2) From this folder: docker compose up -d | |
| # 3) Open http://localhost:8088/ (UI talks to /v1/* on same origin → nginx → host:8080) | |
| # | |
| # Linux: extra_hosts makes host.docker.internal resolve to the host gateway. | |
| services: | |
| edge: | |
| image: nginx:alpine | |
| ports: | |
| - "${EDGE_PORT:-8088}:80" | |
| environment: | |
| LLAMA_HOST: ${LLAMA_HOST:-host.docker.internal} | |
| LLAMA_PORT: ${LLAMA_PORT:-8080} | |
| volumes: | |
| - ./templates:/etc/nginx/templates:ro | |
| - ../dashboard:/usr/share/nginx/html:ro | |
| extra_hosts: | |
| - "host.docker.internal:host-gateway" | |
| restart: unless-stopped | |