"vllm/entrypoints/openai/chat_completion/serving.py" did not exist on "e07aff9e52342dc82b73c803ba69601242801bc4"
agg.yaml 2.91 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeployment
metadata:
  name: sglang-agg
spec:
  services:
    Frontend:
      livenessProbe:
        httpGet:
          path: /health
          port: 8000
15
16
17
18
        initialDelaySeconds: 20
        periodSeconds: 5
        timeoutSeconds: 5
        failureThreshold: 3
19
20
21
22
23
      readinessProbe:
        exec:
          command:
            - /bin/sh
            - -c
24
            - 'curl -s http://localhost:8000/health | jq -e ".status == \"healthy\""'
25
26
27
28
29
30
31
32
33
        initialDelaySeconds: 60
        periodSeconds: 60
        timeoutSeconds: 30
        failureThreshold: 10
      dynamoNamespace: sglang-agg
      componentType: main
      replicas: 1
      resources:
        requests:
34
          cpu: "10"
35
36
          memory: "10Gi"
        limits:
37
38
          cpu: "32"
          memory: "40Gi"
39
40
41
42
43
44
      extraPodSpec:
        mainContainer:
          image: my-registry/sglang-runtime:my-tag
          workingDir: /workspace/components/backends/sglang
          command: ["sh", "-c"]
          args:
45
            - "python3 -m dynamo.sglang.utils.clear_namespace --namespace sglang-agg && python3 -m dynamo.frontend --http-port=8000"
46
47
48
    SGLangDecodeWorker:
      envFromSecret: hf-token-secret
      livenessProbe:
49
50
51
52
        httpGet:
          path: /live
          port: 9090
        periodSeconds: 5
53
        timeoutSeconds: 30
54
        failureThreshold: 1
55
56
      readinessProbe:
        exec:
57
58
59
60
        httpGet:
          path: /health
          port: 9090
        periodSeconds: 10
61
        timeoutSeconds: 30
62
        failureThreshold: 60
63
64
65
66
67
68
69
70
71
      dynamoNamespace: sglang-agg
      componentType: worker
      replicas: 1
      resources:
        requests:
          cpu: "10"
          memory: "20Gi"
          gpu: "1"
        limits:
72
73
          cpu: "32"
          memory: "80Gi"
74
          gpu: "1"
75
76
77
78
79
80
81
      envs:
        - name: DYN_SYSTEM_ENABLED
          value: "true"
        - name: DYN_SYSTEM_USE_ENDPOINT_HEALTH_STATUS
          value: "[\"generate\"]"
        - name: DYN_SYSTEM_PORT
          value: "9090"
82
83
      extraPodSpec:
        mainContainer:
84
85
86
87
88
89
          startupProbe:
            httpGet:
              path: /live
              port: 9090
            periodSeconds: 10
            failureThreshold: 60
90
91
          image: my-registry/sglang-runtime:my-tag
          workingDir: /workspace/components/backends/sglang
92
93
94
          command:
            - /bin/sh
            - -c
95
96
97
98
99
100
101
102
103
104
105
106
107
108
          args:
            - "python3"
            - "-m"
            - "dynamo.sglang.worker"
            - "--model-path"
            - "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
            - "--served-model-name"
            - "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
            - "--page-size"
            - "16"
            - "--tp"
            - "1"
            - "--trust-remote-code"
            - "--skip-tokenizer-init"