agg.yaml 1.02 KB
Newer Older
1
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2
# SPDX-License-Identifier: Apache-2.0
Alec's avatar
Alec committed
3

4
5
6
apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeployment
metadata:
7
  name: vllm-agg
8
9
10
spec:
  services:
    Frontend:
11
      envFromSecret: hf-token-secret
12
      componentType: frontend
13
14
15
      replicas: 1
      extraPodSpec:
        mainContainer:
16
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:my-tag
17
18
    VllmDecodeWorker:
      envFromSecret: hf-token-secret
19
      componentType: worker
20
      replicas: 1
21
22
23
      resources:
        limits:
          gpu: "1"
24
25
26
27
        requests:
          custom:
            # Increase this value for larger models
            ephemeral-storage: "2Gi"
28
29
      extraPodSpec:
        mainContainer:
30
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:my-tag
31
          workingDir: /workspace/examples/backends/vllm
32
          command:
33
34
35
            - python3
            - -m
            - dynamo.vllm
36
          args:
37
38
            - --model
            - Qwen/Qwen3-0.6B