disagg.yaml

# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
Common:
  model: llava-hf/llava-1.5-7b-hf
  block-size: 64
  max-model-len: 4096
  image-token-id: 32000
  num-patches: 576
  kv-transfer-config: '{"kv_connector":"DynamoNixlConnector"}'

Frontend:
  common-configs: [model]

Processor:
  router: round-robin
  prompt-template: "USER: <image>\n<prompt> ASSISTANT:"
  common-configs: [model, block-size]

VllmDecodeWorker:
  remote-prefill: true
  conditional-disagg: true
  max-local-prefill-length: 10
  max-prefill-queue-size: 2
  ServiceArgs:
    workers: 1
    resources:
      gpu: '1'
  common-configs: [model, block-size, image-token-id, max-model-len, num-patches, kv-transfer-config]

VllmPrefillWorker:
  max-num-batched-tokens: 16384
  ServiceArgs:
    workers: 1
    resources:
      gpu: '1'
  common-configs: [model, block-size, image-token-id, max-model-len, num-patches, kv-transfer-config]

VllmEncodeWorker:
  tensor-parallel-size: 1
  router: random
  ServiceArgs:
    workers: 1
    resources:
      gpu: '1'
  common-configs: [model]