From 9254499bc62d5d792fb9b50fdcef78631160f30b Mon Sep 17 00:00:00 2001 From: Marc Blatter Date: Fri, 10 Jul 2026 17:52:45 +0000 Subject: [PATCH] Initial --- compose-vLLM.yaml | 34 ++++++++++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) create mode 100644 compose-vLLM.yaml diff --git a/compose-vLLM.yaml b/compose-vLLM.yaml new file mode 100644 index 0000000..020efc2 --- /dev/null +++ b/compose-vLLM.yaml @@ -0,0 +1,34 @@ +services: + vllm: + image: vllm/vllm-openai:latest + container_name: vllm + restart: unless-stopped + ipc: host # vLLM braucht shared memory für Tensor-Parallelism + + ports: + - "8000:8000" # OpenAI-kompatible API + + volumes: + - /opt/vLLM/models:/root/.cache/huggingface # HF-Model-Cache + - /opt/vLLM/config:/vllm-workspace/config + + environment: + - NVIDIA_VISIBLE_DEVICES=all + - NVIDIA_DRIVER_CAPABILITIES=compute,utility + + deploy: + resources: + reservations: + devices: + - driver: nvidia + count: all + capabilities: [gpu] + + command: > + --model Qwen/Qwen2.5-14B-Instruct-AWQ + --host 0.0.0.0 + --port 8000 + --gpu-memory-utilization 0.90 + --max-model-len 16384 + --enable-auto-tool-choice + --tool-call-parser hermes \ No newline at end of file