#!/bin/bash
#SBATCH --job-name=quick_batch_a100
#SBATCH --account=bfqt-delta-gpu
#SBATCH --partition=gpuA100x4
#SBATCH --gres=gpu:4
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=16
#SBATCH --mem=128G
#SBATCH --time=01:00:00
#SBATCH --output=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/quick_batch_a100-%j.out
#SBATCH --error=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/quick_batch_a100-%j.err

# Quick test: 10 profiles × 5 sessions = 50 sessions on A100
# Tests batch (vanilla) processing while H200 queue is busy

set -e

cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents
source /u/yurenh2/miniforge3/etc/profile.d/conda.sh
conda activate eval

export HF_HOME=/projects/bfqt/users/yurenh2/hf_cache/huggingface
export PYTHONPATH="${PWD}:${PWD}/../src:${PYTHONPATH}"

MODEL_8B="/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/models/llama-3.1-8b-instruct"
PORT_USER=8004
PORT_AGENT=8003

echo "============================================"
echo "Quick Test: Batch Processing on A100"
echo "============================================"
echo "Profiles: 10"
echo "Sessions/profile: 5"
echo "Total: 50 sessions"
echo ""
date
nvidia-smi --query-gpu=index,name,memory.total --format=csv
echo ""

# Kill any existing servers
pkill -f "vllm.entrypoints" 2>/dev/null || true
sleep 2

# Start vLLM servers
echo "Starting 8B user simulator (GPU 0-1, TP=2)..."
CUDA_VISIBLE_DEVICES=0,1 python -m vllm.entrypoints.openai.api_server \
    --model $MODEL_8B \
    --port $PORT_USER \
    --tensor-parallel-size 2 \
    --gpu-memory-utilization 0.85 \
    --max-model-len 4096 \
    --disable-log-requests \
    --dtype bfloat16 &
SERVER_USER_PID=$!

echo "Starting 8B agent (GPU 2-3, TP=2)..."
CUDA_VISIBLE_DEVICES=2,3 python -m vllm.entrypoints.openai.api_server \
    --model $MODEL_8B \
    --port $PORT_AGENT \
    --tensor-parallel-size 2 \
    --gpu-memory-utilization 0.85 \
    --max-model-len 4096 \
    --disable-log-requests \
    --dtype bfloat16 &
SERVER_AGENT_PID=$!

echo "Waiting for servers..."
for i in $(seq 1 100); do
    READY_USER=$(curl -s http://localhost:$PORT_USER/health > /dev/null 2>&1 && echo 1 || echo 0)
    READY_AGENT=$(curl -s http://localhost:$PORT_AGENT/health > /dev/null 2>&1 && echo 1 || echo 0)
    if [ "$READY_USER" = "1" ] && [ "$READY_AGENT" = "1" ]; then
        echo "Both servers ready after $((i*3))s"
        break
    fi
    if [ $((i % 20)) -eq 0 ]; then
        echo "  Still waiting... ($((i*3))s)"
    fi
    sleep 3
done

if ! curl -s http://localhost:$PORT_USER/health > /dev/null; then
    echo "ERROR: User server not healthy"; kill $SERVER_USER_PID $SERVER_AGENT_PID 2>/dev/null; exit 1
fi
if ! curl -s http://localhost:$PORT_AGENT/health > /dev/null; then
    echo "ERROR: Agent server not healthy"; kill $SERVER_USER_PID $SERVER_AGENT_PID 2>/dev/null; exit 1
fi
echo "Both servers healthy"
echo ""

# Run quick test with vanilla (batch)
echo "============================================"
echo "Test: BATCH processing (vanilla method)"
echo "============================================"
START=$(date +%s)

PROFILE_PATH="/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/data/complex_profiles_v2/profiles_200.jsonl"

python scripts/run_experiments.py \
    --methods vanilla \
    --datasets mmlu \
    --n-profiles 10 \
    --n-sessions 5 \
    --use-vllm \
    --batch-size 50 \
    --parallel-profiles 10 \
    --output-dir ../results/quick_test_batch_a100 \
    --profile-path "$PROFILE_PATH"

END=$(date +%s)
ELAPSED=$((END-START))
echo ""
echo "Vanilla (batch) completed in ${ELAPSED}s"

# Cleanup
echo ""
echo "Cleaning up..."
kill $SERVER_USER_PID $SERVER_AGENT_PID 2>/dev/null || true

echo ""
echo "============================================"
echo "QUICK TEST RESULTS (A100)"
echo "============================================"
echo ""
echo "Vanilla (BATCH): ${ELAPSED}s for 50 sessions"
echo ""

if [ $ELAPSED -gt 0 ]; then
    THROUGHPUT=$((50 * 3600 / ELAPSED))
    echo "Throughput: ${THROUGHPUT} sessions/hr"
fi

echo ""
echo "Results saved to: ../results/quick_test_batch_a100/"
echo ""
date