#!/bin/bash
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
set -e
trap 'echo Cleaning up...; kill 0' EXIT

# Common configuration
MODEL="Qwen/Qwen3-30B-A3B"

HTTP_PORT="${DYN_HTTP_PORT:-8000}"
echo "=========================================="
echo "Launching Data Parallel / Expert Parallelism (4 GPUs)"
echo "=========================================="
echo "Model:       $MODEL"
echo "Frontend:    http://localhost:$HTTP_PORT"
echo "=========================================="
echo ""
echo "Example test command:"
echo ""
echo "  curl http://localhost:${HTTP_PORT}/v1/chat/completions \\"
echo "    -H 'Content-Type: application/json' \\"
echo "    -d '{"
echo "      \"model\": \"${MODEL}\","
echo "      \"messages\": [{\"role\": \"user\", \"content\": \"Explain why Roger Federer is considered one of the greatest tennis players of all time\"}],"
echo "      \"max_tokens\": 32"
echo "    }'"
echo ""
echo "=========================================="

# run ingress
# dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
python -m dynamo.frontend --router-mode kv &

# Data Parallel Attention / Expert Parallelism
# Routing to DP workers managed by Dynamo
# Chose Qwen3-30B because its a small MOE that can fit on smaller GPUs (L40S for example)
# --enforce-eager is added for quick deployment. for production use, need to remove this flag
for i in {0..3}; do
    VLLM_NIXL_SIDE_CHANNEL_PORT=$((20096 + i)) \
    CUDA_VISIBLE_DEVICES=$i python3 -m dynamo.vllm \
    --model "$MODEL" \
    --data-parallel-rank $i \
    --data-parallel-size 4 \
    --enable-expert-parallel \
    --enforce-eager \
    --kv-events-config "{\"publisher\":\"zmq\",\"topic\":\"kv-events\",\"endpoint\":\"tcp://*:$((20080 + i))\",\"enable_kv_cache_events\":true}" &
done

echo "All workers starting. (press Ctrl+C to stop)..."
wait