Training updates

genesis/docker-compose.network.yml

@@ -35,6 +35,7 @@ services:
       - DAPI_API__PUBLIC_SERVER_PORT=9000
       - DAPI_API__ML_SERVER_PORT=9100
       - DAPI_API__ADMIN_SERVER_PORT=9200
+      - DAPI_API__ML_GRPC_SERVER_PORT=9300
       - NODE_CONFIG_PATH=/root/node_config.json
     volumes:
       - /srv/dai/inference:/root/.inference
@@ -43,6 +44,7 @@ services:
       - "8000:9000"
       - "9100:9100"
       - "9200:9200"
+      - "9300:9300"
 
 networks:
   genesis-network:

genesis/node-config.json

@@ -12,7 +12,13 @@
           "fp8"
         ]
       }
-    }
+    },
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
   },
   {
     "id": "mlnode2",
@@ -29,6 +35,12 @@
           "fp8"
         ]
       }
-    }
+    },
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
   }
-]
+]

join-2/config.env

@@ -0,0 +1,7 @@
+export VERSION="0.0.1-training-3"
+export KEY_NAME=node1
+export PUBLIC_URL=http://172.18.114.144:8000
+export SEED_API_URL=http://172.18.114.148:8000
+export SEED_NODE_RPC_URL=http://172.18.114.148:26657
+export SEED_NODE_P2P_URL=http://172.18.114.148:26656
+export PORT=8000

join-2/docker-compose-cloud-join.yml

@@ -0,0 +1,62 @@
+services:
+  node:
+    container_name: node
+    image: 172.18.114.101:5556/decentralized-ai/inferenced:${VERSION:-latest}
+    command: ["sh", "./init-docker.sh"]
+    volumes:
+      - .inference:/root/.inference
+    environment:
+      - SEED_NODE_RPC_URL=${SEED_NODE_RPC_URL}
+      - SEED_NODE_P2P_URL=${SEED_NODE_P2P_URL}
+      - SYNC_WITH_SNAPSHOTS=${SYNC_WITH_SNAPSHOTS} # sync using snapshots?
+      - RPC_SERVER_URL_2=${RPC_SERVER_URL_2}       # rpc-servers to get snapshots
+      - RPC_SERVER_URL_1=${RPC_SERVER_URL_1}
+      - TRUSTED_BLOCK_PERIOD=${TRUSTED_BLOCK_PERIOD:-2000} # num of blocks to return back from latest block to get the earliest trusted snapshot
+      - KEY_NAME=${KEY_NAME}
+    ports:
+      - "26656:26656" #p2p
+      - "26657:26657" #rpc
+  api:
+    container_name: api
+    image: 172.18.114.101:5556/decentralized-ai/api:${VERSION:-latest}
+    volumes:
+      - .inference:/root/.inference
+      - ./node-config.json:/root/node_config.json
+    depends_on:
+      - node
+    environment:
+      - KEY_NAME=${KEY_NAME}
+      - DAPI_API__POC_CALLBACK_URL=http://api:8080
+      - DAPI_API__PUBLIC_URL=${PUBLIC_URL}
+      - DAPI_CHAIN_NODE__SEED_API_URL=${SEED_API_URL}
+      - NODE_CONFIG_PATH=/root/node_config.json
+    ports:
+      - "${PORT:-8080}:8080"
+    restart: always
+
+  inference-node:
+    image: 172.18.114.101:5556/decentralized-ai/mlnode:3.0.1-alpha1
+    volumes:
+      - ${HOME}/cache:/root/.cache
+    environment:
+      - HF_HOME=/root/.cache
+      #- HF_HUB_ENABLE_HF_TRANSFER=true
+      - HF_TOKEN=${HF_TOKEN}
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
+    command: >
+      uvicorn api.app:app
+      --host=0.0.0.0
+      --port=8080
+    ports:
+      - "8080:8080"
+      - "5000:5000"
+    restart: always
+
+volumes:
+  inference-data:

join-2/node-config.json

@@ -0,0 +1,18 @@
+[
+  {
+    "id": "node1",
+    "host": "172.18.114.144",
+    "inference_port": 5000,
+    "poc_port": 8080,
+    "max_concurrent": 500,
+    "models": [
+      "unsloth/llama-3-8b-Instruct"
+    ],
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
+  }
+]

join/docker-compose.yml

@@ -53,10 +53,12 @@ services:
       - DAPI_API__PUBLIC_SERVER_PORT=9000
       - DAPI_API__ML_SERVER_PORT=9100
       - DAPI_API__ADMIN_SERVER_PORT=9200
+      - DAPI_API__ML_GRPC_SERVER_PORT=9300
     ports:
       - "${API_PORT}:9000"
       - "9100:9100"
       - "9200:9200"
+      - "9300:9300"
     restart: always
 
   inference:

join/node-config.json

@@ -12,6 +12,12 @@
           "fp8"
         ]
       }
-    }
+    },
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
   }
 ]

test/genesis/docker-compose.network.yml

@@ -36,13 +36,15 @@ services:
       - DAPI_API__PUBLIC_SERVER_PORT=9000
       - DAPI_API__ML_SERVER_PORT=9100
       - DAPI_API__ADMIN_SERVER_PORT=9200
+      - DAPI_API__ML_GRPC_SERVER_PORT=9300
     volumes:
       - /srv/dai/inference:/root/.inference
       - ./node-config.json:/root/node_config.json
     ports:
       - "8000:9000"
       - "9100:9100"
       - "9200:9200"
+      - "9300:9300"
 
   inference:
     image: ghcr.io/product-science/mlnode:3.0.4-alpha2

test/genesis/node-config.json

@@ -12,6 +12,12 @@
           "fp8"
         ]
       }
-    }
+    },
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
   }
 ]

test/join/docker-compose.yml

@@ -51,10 +51,12 @@ services:
       - DAPI_API__PUBLIC_SERVER_PORT=9000
       - DAPI_API__ML_SERVER_PORT=9100
       - DAPI_API__ADMIN_SERVER_PORT=9200
+      - DAPI_API__ML_GRPC_SERVER_PORT=9300
     ports:
       - "${API_PORT}:9000"
       - "9100:9100"
       - "9200:9200"
+      - "9300:9300"
     restart: always
 
   inference:
@@ -64,6 +66,7 @@ services:
     environment:
       - HF_HOME=/root/.cache
       - HF_HUB_ENABLE_HF_TRANSFER=true
+      - HF_TOKEN=${HF_TOKEN}
     deploy:
       resources:
         reservations:
@@ -77,6 +80,8 @@ services:
       uvicorn api.app:app
       --host=0.0.0.0
       --port=8080
+    ports:
+      - "8080:8080"
     restart: always
 
 volumes:

test/join/node-config.json

@@ -12,6 +12,12 @@
           "fp8"
         ]
       }
-    }
+    },
+    "hardware": [
+      {
+        "type": "3090",
+        "count": 1
+      }
+    ]
   }
 ]