disagg_router.yaml 1.83 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
# 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: trtllm-v1-disagg-router
spec:
  services:
    Frontend:
      dynamoNamespace: trtllm-v1-disagg-router
12
      componentType: frontend
13
14
15
      replicas: 1
      extraPodSpec:
        mainContainer:
16
          image: my-registry/trtllm-runtime:my-tag
17
18
19
      envs:
        - name: DYN_ROUTER_MODE
          value: kv
20
21
22
23
24
25
26
27
28
29
    TRTLLMPrefillWorker:
      dynamoNamespace: trtllm-v1-disagg-router
      envFromSecret: hf-token-secret
      componentType: worker
      replicas: 2
      resources:
        limits:
          gpu: "1"
      extraPodSpec:
        mainContainer:
30
          image: my-registry/trtllm-runtime:my-tag
31
32
33
34
35
          workingDir: /workspace/components/backends/trtllm
          command:
            - /bin/sh
            - -c
          args:
36
            - "python3 -m dynamo.trtllm --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --extra-engine-args engine_configs/prefill.yaml --disaggregation-mode prefill --disaggregation-strategy prefill_first --publish-events-and-metrics"
37
38
39
40
41
42
43
44
45
46
    TRTLLMDecodeWorker:
      dynamoNamespace: trtllm-v1-disagg-router
      envFromSecret: hf-token-secret
      componentType: worker
      replicas: 1
      resources:
        limits:
          gpu: "1"
      extraPodSpec:
        mainContainer:
47
          image: my-registry/trtllm-runtime:my-tag
48
49
50
51
52
          workingDir: /workspace/components/backends/trtllm
          command:
            - /bin/sh
            - -c
          args:
53
            - "python3 -m dynamo.trtllm --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --extra-engine-args engine_configs/decode.yaml --disaggregation-mode decode --disaggregation-strategy prefill_first"