awesome-akash/Qwen3-235B-A22B-FP8/deploy.yaml at master · akash-network/awesome-akash · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
---
version: "2.0"
services:
  sglang:
    image: lmsysorg/sglang:v0.4.6.post1-cu125-srt
    expose:
      - port: 8000
        as: 8000
        to:
          - global: true
    command:
      - bash
      - "-c"
    args:
      - >-
        python3 -m sglang.launch_server --model-path Qwen/Qwen3-235B-A22B-FP8 --tensor-parallel-size 4 --host 0.0.0.0 --port 8000 --reasoning-parser qwen3 --json-model-override-args '{"rope_scaling":{"rope_type":"yarn","factor":4.0,"original_max_position_embeddings":32768}}'
    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:
    sglang:
      resources:
        cpu:
          units: 64
        memory:
          size: 128Gi
        storage:
          - size: 50Gi
          - name: data
            size: 300Gi
            attributes:
              persistent: true
              class: beta3
          - name: shm
            size: 10Gi
            attributes:
              class: ram
              persistent: false
        gpu:
          units: 4
          attributes:
            vendor:
              nvidia:
                - model: h100
                  ram: 80Gi
                - model: a100
                  ram: 80Gi
  placement:
    dcloud:
      pricing:
        sglang:
          denom: uakt
          amount: 1000000
deployment:
  sglang:
    dcloud:
      profile: sglang
      count: 1