deploy.yaml

---
version: "2.0"
services:
  sglang1:
    image: lmsysorg/sglang:v0.4.1.post4-cu124-srt
    expose:
      - port: 8000
        as: 8000
        to:
          - global: true
    command:
      - bash
      - "-c"
    args:
      - >- 
        python3 -m sglang.launch_server --model deepseek-ai/DeepSeek-R1-Distill-Llama-70B --tp 2 --trust-remote-code --host 0.0.0.0 --port 8000
    params:
      storage:
        shm:
          mount: /dev/shm
        data:
          mount: /root/.cache # Mount the data storage to the cache directory for persistent storage of model files
          readOnly: false
profiles:
  compute:
    sglang1:
      resources:
        cpu:
          units: 32
        memory:
          size: 128Gi
        storage:
          - size: 50Gi
          - name: data
            size: 200Gi
            attributes:
              persistent: true
              class: beta3
          - name: shm
            size: 32Gi
            attributes:
              class: ram
              persistent: false
        gpu:
          units: 2
          attributes:
            vendor:
              nvidia:
                - model: a100
                - model: h100
  placement:
    dcloud:
      pricing:
        sglang1:
          denom: uakt
          amount: 1000000
deployment:
  sglang1:
    dcloud:
      profile: sglang1
      count: 1