diff --git a/docker-compose.yml b/docker-compose.yml new file mode 100644 index 0000000..1556af2 --- /dev/null +++ b/docker-compose.yml @@ -0,0 +1,53 @@ +version: "3" + +services: + + beluga: + image: learningathome/petals:main + depends_on: + - backbone + environment: +# - PJRT_DEVICE=TPU +#--device xla + - BOOTSTRAP_INFO=/ip4/0.0.0.0/tcp/8099 + command: python -m petals.cli.run_server --port 31330 --num_blocks=1 petals-team/StableBeluga2 --initial_peers /ip4/192.168.1.90/tcp/8099/p2p/QmfVvYv3w3EqpKGYG5FCcER9bFgoGLCUvXDUJsZAgSDw3N + + deploy: + resources: + reservations: + devices: + - driver: nvidia + count: 1 + capabilities: [gpu] + ports: + - "31330:31330" + restart: always + + backbone: + image: learningathome/petals:main + command: python -m petals.cli.run_dht --host_maddrs /ip4/0.0.0.0/tcp/8099 --identity_path /cache/bootstrap1.id + volumes: + - petals-cache-backbone:/cache + network_mode: host + ipc: host + restart: unless-stopped + + +# version: '3.8' +# services: +# backbone: +# image: learningathome/petals:main +# environment: +# - BOOTSTRAP_INFO=/ip4/0.0.0.0/tcp/8099 +# command: python -m petals.cli.run_dht --host_maddrs ${BOOTSTRAP_INFO} --identity_path bootstrap1.id +# volumes: +# - type: volume +# source: petals-cache-backbone +# target: /cache +# network_mode: host +# ipc: host +# restart: always +# container_name: backbone + +volumes: + petals-cache-backbone: diff --git a/get_peersl.sh b/get_peersl.sh new file mode 100644 index 0000000..5dcd3ab --- /dev/null +++ b/get_peersl.sh @@ -0,0 +1 @@ +docker logs petals-backbone-1 2>&1 |grep initial_peers |cut "-d " -f18- | sort -u > peers.txt