Update vllm compose and run YAMLs

distributions/remote-vllm/compose.yaml

@@ -1,43 +1,83 @@
+# NOTES:
+#
+# This Docker Compose (and the associated run.yaml) assumes you will be
+# running in the default "bridged" network mode.
+#
+# If you need "host" network mode, please uncomment
+#  - network_mode: "host"
+# and comment the lines with port mapping
+#  - ports:
+#    - "5100:5100"
+#
+# Similarly change "host.docker.internal" to "localhost" in the run.yaml file
+#
 services:
-  vllm:
+  vllm-0:
     image: vllm/vllm-openai:latest
-    network_mode: "host"
     volumes:
       - $HOME/.cache/huggingface:/root/.cache/huggingface
+    # network_mode: "host"
     ports:
-      - "8000:8000"
+       - "5100:5100"
     devices:
       - nvidia.com/gpu=all
     environment:
-      - CUDA_VISIBLE_DEVICES=0
-    command: []
+      - CUDA_VISIBLE_DEVICES=4
+      - HUGGING_FACE_HUB_TOKEN=$HF_TOKEN
+    command: >
+      --gpu-memory-utilization 0.75
+      --model meta-llama/Llama-3.1-8B-Instruct
+      --enforce-eager
+      --max-model-len 8192
+      --max-num-seqs 16
+      --port 5100
+    deploy:
+      resources:
+        reservations:
+          devices:
+          - driver: nvidia
+            capabilities: [gpu]
+    runtime: nvidia
+  vllm-1:
+    image: vllm/vllm-openai:latest
+    volumes:
+      - $HOME/.cache/huggingface:/root/.cache/huggingface
+    # network_mode: "host"
+    ports:
+      - "5101:5101"
+    devices:
+      - nvidia.com/gpu=all
+    environment:
+      - CUDA_VISIBLE_DEVICES=5
+      - HUGGING_FACE_HUB_TOKEN=$HF_TOKEN
+    command: >
+      --gpu-memory-utilization 0.75
+      --model meta-llama/Llama-Guard-3-1B
+      --enforce-eager
+      --max-model-len 8192
+      --max-num-seqs 16
+      --port 5101
     deploy:
       resources:
         reservations:
           devices:
           - driver: nvidia
-            # that's the closest analogue to --gpus; provide
-            # an integer amount of devices or 'all'
-            count: 1
-            # Devices are reserved using a list of capabilities, making
-            # capabilities the only required field. A device MUST
-            # satisfy all the requested capabilities for a successful
-            # reservation.
             capabilities: [gpu]
     runtime: nvidia
   llamastack:
     depends_on:
-    - vllm
-    image: llamastack/distribution-remote-vllm
-    network_mode: "host"
+    - vllm-0
+    - vllm-1
+      # image: llamastack/distribution-remote-vllm
+    image: localhost/distribution-remote-vllm:test-0.0.52rc3
     volumes:
       - ~/.llama:/root/.llama
-      # Link to ollama run.yaml file
-      - ./run.yaml:/root/llamastack-run-remote-vllm.yaml
+      - ~/local/llama-stack/distributions/remote-vllm/run.yaml:/root/llamastack-run-remote-vllm.yaml
+    # network_mode: "host"
     ports:
-      - "5000:5000"
-    # Hack: wait for vllm server to start before starting docker
-    entrypoint: bash -c "sleep 60; python -m llama_stack.distribution.server.server --yaml_config /root/llamastack-run-remote-vllm.yaml"
+      - "5001:5001"
+    # Hack: wait for vLLM server to start before starting docker
+    entrypoint: bash -c "sleep 60; python -m llama_stack.distribution.server.server --yaml_config /root/llamastack-run-remote-vllm.yaml --port 5001"
     deploy:
       restart_policy:
         condition: on-failure
@@ -45,4 +85,6 @@ services:
         max_attempts: 5
         window: 60s
 volumes:
-  vllm:
+  vllm-0:
+  vllm-1:
+  llamastack: