#!/bin/bash
set -e

echo "========================================="
echo "DQN Hyperopt Deployment (CORRECTED OBJECTIVE)"
echo "========================================="
echo ""

# Configuration
TIMESTAMP=$(date +%Y%m%d_%H%M%S)
OUTPUT_DIR="dqn_hyperopt_corrected_${TIMESTAMP}"

echo "Configuration:"
echo "  Objective: Episode rewards (CORRECTED from validation loss)"
echo "  Trials: 50"
echo "  Epochs per trial: 100"
echo "  GPU: RTX A4000 ($0.25/hr)"
echo "  Expected duration: 12-25 min"
echo "  Expected cost: $0.05-$0.10"
echo "  Output: /runpod-volume/ml_training/${OUTPUT_DIR}"
echo ""

# Verify Docker image is up-to-date
echo "Verifying Docker image..."
IMAGE_DATE=$(docker images jgrusewski/foxhunt-hyperopt:latest --format "{{.CreatedAt}}" | head -1)
echo "  Image timestamp: ${IMAGE_DATE}"
echo "  Expected: Nov 1, 2025 23:18+ (after fix)"
echo ""

# Deploy DQN hyperopt with CORRECTED objective
echo "Deploying DQN hyperopt pod..."
python3 scripts/python/runpod/runpod_deploy.py \
    --gpu-type "RTX A4000" \
    --image "jgrusewski/foxhunt-hyperopt:latest" \
    --command "hyperopt_dqn_demo --parquet-file /runpod-volume/test_data/ES_FUT_180d.parquet --trials 50 --epochs 100 --base-dir /runpod-volume/ml_training/${OUTPUT_DIR}"

echo ""
echo "✅ DQN hyperopt deployment initiated (CORRECTED OBJECTIVE)"
echo "Monitor logs: python3 scripts/python/runpod/monitor_logs.py <pod_id>"
echo ""
echo "CRITICAL FIX APPLIED:"
echo "  Previous objective: val_loss (WRONG - rewarded tiny batches)"
echo "  New objective: -avg_episode_reward (CORRECT)"
echo ""
echo "Expected Results:"
echo "  Batch size: Should vary widely (not stuck at 32-43)"
echo "  Learning rate: Should optimize for actual learning"
echo "  Episode rewards: Should maximize trading returns"
echo "  Q-values: Should show proper value estimation"
echo ""
echo "Previous Issue (FIXED):"
echo "  Tiny batch sizes (32-43) prevented learning"
echo "  Q-values stayed near zero (noisy gradients)"
echo "  Validation loss was artificially low (misleading)"
echo ""
echo "Next Steps:"
echo "  1. Monitor pod logs for trial progress"
echo "  2. Check best hyperparameters after completion"
echo "  3. Compare batch sizes to previous run (32-43)"
echo "  4. Verify Q-values and episode rewards improve"
echo ""