disagg_multimodal.sh

#!/bin/bash
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

# Environment variables with defaults
export DYNAMO_HOME=${DYNAMO_HOME:-"/workspace"}
export MODEL_PATH=${MODEL_PATH:-"Qwen/Qwen2-VL-7B-Instruct"}
export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen/Qwen2-VL-7B-Instruct"}
export PREFILL_ENGINE_ARGS=${PREFILL_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen2-vl-7b-instruct/prefill.yaml"}
export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen2-vl-7b-instruct/decode.yaml"}
export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"}
export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"}
export MODALITY=${MODALITY:-"multimodal"}

# Setup cleanup trap
cleanup() {
    echo "Cleaning up background processes..."
    kill $DYNAMO_PID $PREFILL_PID 2>/dev/null || true
    wait $DYNAMO_PID $PREFILL_PID 2>/dev/null || true
    echo "Cleanup complete."
}
trap cleanup EXIT INT TERM


# run frontend
# dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
python3 -m dynamo.frontend &
DYNAMO_PID=$!

# run prefill worker
CUDA_VISIBLE_DEVICES=$PREFILL_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
  --model-path "$MODEL_PATH" \
  --served-model-name "$SERVED_MODEL_NAME" \
  --extra-engine-args  "$PREFILL_ENGINE_ARGS" \
  --modality "$MODALITY" \
  --disaggregation-mode prefill &
PREFILL_PID=$!

# run decode worker
CUDA_VISIBLE_DEVICES=$DECODE_CUDA_VISIBLE_DEVICES python3 -m dynamo.trtllm \
  --model-path "$MODEL_PATH" \
  --served-model-name "$SERVED_MODEL_NAME" \
  --extra-engine-args  "$DECODE_ENGINE_ARGS" \
  --modality "$MODALITY" \
  --disaggregation-mode decode