555 lines
22 KiB
Python
555 lines
22 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Test Script for Real-time RL COB Trader
|
|
|
|
This script tests the real-time reinforcement learning system to ensure:
|
|
1. Proper model initialization and parameter count (~1B parameters)
|
|
2. COB data integration and feature extraction
|
|
3. Real-time inference pipeline
|
|
4. Signal accumulation and consensus
|
|
5. Training loop functionality
|
|
6. Trade execution integration
|
|
|
|
Run this before deploying the live system.
|
|
"""
|
|
|
|
import asyncio
|
|
import logging
|
|
import numpy as np
|
|
import torch
|
|
import time
|
|
import json
|
|
from datetime import datetime
|
|
from typing import Dict, Any
|
|
|
|
# Local imports
|
|
from core.realtime_rl_cob_trader import RealtimeRLCOBTrader, MassiveRLNetwork, PredictionResult
|
|
from core.trading_executor import TradingExecutor
|
|
|
|
# Configure logging
|
|
logging.basicConfig(
|
|
level=logging.INFO,
|
|
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
|
|
)
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
class RealtimeRLTester:
|
|
"""
|
|
Comprehensive tester for Real-time RL COB Trader
|
|
"""
|
|
|
|
def __init__(self):
|
|
self.test_results = {}
|
|
self.trader = None
|
|
self.trading_executor = None
|
|
|
|
async def run_all_tests(self):
|
|
"""Run all tests and generate report"""
|
|
logger.info("=" * 60)
|
|
logger.info("REAL-TIME RL COB TRADER TESTING SUITE")
|
|
logger.info("=" * 60)
|
|
|
|
tests = [
|
|
self.test_model_initialization,
|
|
self.test_model_parameter_count,
|
|
self.test_feature_extraction,
|
|
self.test_inference_performance,
|
|
self.test_signal_accumulation,
|
|
self.test_training_pipeline,
|
|
self.test_trading_integration,
|
|
self.test_performance_monitoring
|
|
]
|
|
|
|
for test in tests:
|
|
try:
|
|
await test()
|
|
except Exception as e:
|
|
logger.error(f"Test {test.__name__} failed: {e}")
|
|
self.test_results[test.__name__] = {'status': 'FAILED', 'error': str(e)}
|
|
|
|
await self.generate_test_report()
|
|
|
|
async def test_model_initialization(self):
|
|
"""Test model initialization and architecture"""
|
|
logger.info("🧠 Testing Model Initialization...")
|
|
|
|
try:
|
|
# Test model creation
|
|
model = MassiveRLNetwork(input_size=2000, hidden_size=4096, num_layers=12)
|
|
|
|
# Check if CUDA is available
|
|
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
|
model = model.to(device)
|
|
|
|
# Test forward pass
|
|
batch_size = 4
|
|
test_input = torch.randn(batch_size, 2000).to(device)
|
|
|
|
with torch.no_grad():
|
|
outputs = model(test_input)
|
|
|
|
# Verify outputs
|
|
assert 'price_logits' in outputs
|
|
assert 'value' in outputs
|
|
assert 'confidence' in outputs
|
|
assert 'features' in outputs
|
|
|
|
assert outputs['price_logits'].shape == (batch_size, 3) # DOWN, SIDEWAYS, UP
|
|
assert outputs['value'].shape == (batch_size, 1)
|
|
assert outputs['confidence'].shape == (batch_size, 1)
|
|
|
|
self.test_results['test_model_initialization'] = {
|
|
'status': 'PASSED',
|
|
'device': str(device),
|
|
'output_shapes': {k: list(v.shape) for k, v in outputs.items()}
|
|
}
|
|
|
|
logger.info("✅ Model initialization test PASSED")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_model_initialization'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_model_parameter_count(self):
|
|
"""Test that model has approximately 400M parameters"""
|
|
logger.info("🔢 Testing Model Parameter Count...")
|
|
|
|
try:
|
|
model = MassiveRLNetwork(input_size=2000, hidden_size=2048, num_layers=8)
|
|
|
|
total_params = sum(p.numel() for p in model.parameters())
|
|
trainable_params = sum(p.numel() for p in model.parameters() if p.requires_grad)
|
|
|
|
logger.info(f"Total parameters: {total_params:,}")
|
|
logger.info(f"Trainable parameters: {trainable_params:,}")
|
|
|
|
# Check if parameters are approximately 400M (350M - 450M range)
|
|
target_400m = total_params >= 350_000_000 and total_params <= 450_000_000
|
|
|
|
self.test_results['test_model_parameter_count'] = {
|
|
'status': 'PASSED' if target_400m else 'WARNING',
|
|
'total_parameters': total_params,
|
|
'trainable_parameters': trainable_params,
|
|
'parameter_size_gb': (total_params * 4) / (1024**3), # 4 bytes per float32
|
|
'is_optimized': target_400m, # Around 400M parameters for faster startup
|
|
'target_range': '350M - 450M parameters'
|
|
}
|
|
|
|
logger.info(f"✅ Model has {total_params:,} parameters ({total_params/1e6:.0f}M)")
|
|
if target_400m:
|
|
logger.info("✅ Parameter count within 400M target range for fast startup")
|
|
else:
|
|
logger.warning(f"⚠️ Parameter count outside 400M target range: {total_params/1e6:.0f}M")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_model_parameter_count'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_feature_extraction(self):
|
|
"""Test feature extraction from COB data"""
|
|
logger.info("🔍 Testing Feature Extraction...")
|
|
|
|
try:
|
|
# Initialize trader
|
|
self.trading_executor = TradingExecutor(simulation_mode=True)
|
|
self.trader = RealtimeRLCOBTrader(
|
|
symbols=['BTC/USDT'],
|
|
trading_executor=self.trading_executor,
|
|
inference_interval_ms=1000 # Slower for testing
|
|
)
|
|
|
|
# Create mock COB data
|
|
mock_cob_data = {
|
|
'state': np.random.randn(1500), # Mock state features
|
|
'timestamp': datetime.now(),
|
|
'type': 'cob_state'
|
|
}
|
|
|
|
# Test feature extraction
|
|
features = self.trader._extract_features('BTC/USDT', mock_cob_data)
|
|
|
|
assert features is not None
|
|
assert len(features) == 2000 # Target feature size
|
|
assert features.dtype == np.float32
|
|
assert not np.any(np.isnan(features))
|
|
assert not np.any(np.isinf(features))
|
|
|
|
# Test normalization
|
|
assert np.abs(np.mean(features)) < 1.0 # Roughly normalized
|
|
assert np.std(features) < 10.0 # Not too spread out
|
|
|
|
self.test_results['test_feature_extraction'] = {
|
|
'status': 'PASSED',
|
|
'feature_size': len(features),
|
|
'feature_range': [float(np.min(features)), float(np.max(features))],
|
|
'feature_stats': {
|
|
'mean': float(np.mean(features)),
|
|
'std': float(np.std(features)),
|
|
'median': float(np.median(features))
|
|
}
|
|
}
|
|
|
|
logger.info("✅ Feature extraction test PASSED")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_feature_extraction'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_inference_performance(self):
|
|
"""Test inference speed and quality"""
|
|
logger.info("⚡ Testing Inference Performance...")
|
|
|
|
try:
|
|
if not self.trader:
|
|
self.trading_executor = TradingExecutor(simulation_mode=True)
|
|
self.trader = RealtimeRLCOBTrader(
|
|
symbols=['BTC/USDT'],
|
|
trading_executor=self.trading_executor
|
|
)
|
|
|
|
# Test multiple inferences
|
|
num_tests = 10
|
|
inference_times = []
|
|
|
|
for i in range(num_tests):
|
|
# Create test features
|
|
test_features = np.random.randn(2000).astype(np.float32)
|
|
test_features = self.trader._normalize_features(test_features)
|
|
|
|
# Time inference
|
|
start_time = time.time()
|
|
prediction = self.trader._predict('BTC/USDT', test_features)
|
|
inference_time = (time.time() - start_time) * 1000
|
|
|
|
inference_times.append(inference_time)
|
|
|
|
# Verify prediction structure
|
|
assert 'direction' in prediction
|
|
assert 'confidence' in prediction
|
|
assert 'change' in prediction
|
|
assert 'value' in prediction
|
|
|
|
assert 0 <= prediction['direction'] <= 2
|
|
assert 0.0 <= prediction['confidence'] <= 1.0
|
|
assert isinstance(prediction['change'], float)
|
|
assert isinstance(prediction['value'], float)
|
|
|
|
avg_inference_time = np.mean(inference_times)
|
|
max_inference_time = np.max(inference_times)
|
|
|
|
# Check if inference is fast enough (target: <50ms per inference)
|
|
inference_target_ms = 50.0
|
|
|
|
self.test_results['test_inference_performance'] = {
|
|
'status': 'PASSED' if avg_inference_time < inference_target_ms else 'WARNING',
|
|
'average_inference_time_ms': float(avg_inference_time),
|
|
'max_inference_time_ms': float(max_inference_time),
|
|
'target_time_ms': inference_target_ms,
|
|
'meets_target': avg_inference_time < inference_target_ms,
|
|
'inferences_per_second': 1000.0 / avg_inference_time
|
|
}
|
|
|
|
logger.info(f"✅ Average inference time: {avg_inference_time:.2f}ms")
|
|
logger.info(f"✅ Max inference time: {max_inference_time:.2f}ms")
|
|
logger.info(f"✅ Inferences per second: {1000.0/avg_inference_time:.1f}")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_inference_performance'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_signal_accumulation(self):
|
|
"""Test signal accumulation and consensus logic"""
|
|
logger.info("🎯 Testing Signal Accumulation...")
|
|
|
|
try:
|
|
if not self.trader:
|
|
self.trading_executor = TradingExecutor(simulation_mode=True)
|
|
self.trader = RealtimeRLCOBTrader(
|
|
symbols=['BTC/USDT'],
|
|
trading_executor=self.trading_executor,
|
|
required_confident_predictions=3
|
|
)
|
|
|
|
symbol = 'BTC/USDT'
|
|
accumulator = self.trader.signal_accumulators[symbol]
|
|
|
|
# Test adding signals
|
|
test_predictions = []
|
|
for i in range(5):
|
|
prediction = PredictionResult(
|
|
timestamp=datetime.now(),
|
|
symbol=symbol,
|
|
predicted_direction=2, # UP
|
|
confidence=0.8,
|
|
predicted_change=0.001,
|
|
features=np.random.randn(2000).astype(np.float32)
|
|
)
|
|
test_predictions.append(prediction)
|
|
self.trader._add_signal(symbol, prediction)
|
|
|
|
# Check accumulator state
|
|
assert len(accumulator.signals) == 5
|
|
assert accumulator.confidence_sum == 5 * 0.8
|
|
assert accumulator.total_predictions == 5
|
|
|
|
# Test consensus logic (simulate processing)
|
|
recent_signals = list(accumulator.signals)[-3:]
|
|
directions = [signal.predicted_direction for signal in recent_signals]
|
|
|
|
# All should be direction 2 (UP)
|
|
direction_counts = {0: 0, 1: 0, 2: 0}
|
|
for direction in directions:
|
|
direction_counts[direction] += 1
|
|
|
|
dominant_direction = max(direction_counts, key=direction_counts.get)
|
|
consensus_count = direction_counts[dominant_direction]
|
|
|
|
assert dominant_direction == 2
|
|
assert consensus_count == 3
|
|
|
|
self.test_results['test_signal_accumulation'] = {
|
|
'status': 'PASSED',
|
|
'signals_added': len(accumulator.signals),
|
|
'confidence_sum': accumulator.confidence_sum,
|
|
'consensus_direction': dominant_direction,
|
|
'consensus_count': consensus_count
|
|
}
|
|
|
|
logger.info("✅ Signal accumulation test PASSED")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_signal_accumulation'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_training_pipeline(self):
|
|
"""Test training pipeline functionality"""
|
|
logger.info("🧠 Testing Training Pipeline...")
|
|
|
|
try:
|
|
if not self.trader:
|
|
self.trading_executor = TradingExecutor(simulation_mode=True)
|
|
self.trader = RealtimeRLCOBTrader(
|
|
symbols=['BTC/USDT'],
|
|
trading_executor=self.trading_executor
|
|
)
|
|
|
|
symbol = 'BTC/USDT'
|
|
|
|
# Create mock training data
|
|
test_predictions = []
|
|
for i in range(10):
|
|
prediction = PredictionResult(
|
|
timestamp=datetime.now(),
|
|
symbol=symbol,
|
|
predicted_direction=np.random.randint(0, 3),
|
|
confidence=np.random.uniform(0.5, 1.0),
|
|
predicted_change=np.random.uniform(-0.001, 0.001),
|
|
features=np.random.randn(2000).astype(np.float32),
|
|
actual_direction=np.random.randint(0, 3),
|
|
actual_change=np.random.uniform(-0.001, 0.001),
|
|
reward=np.random.uniform(-1.0, 1.0)
|
|
)
|
|
test_predictions.append(prediction)
|
|
|
|
# Test training batch
|
|
loss = await self.trader._train_batch(symbol, test_predictions)
|
|
|
|
assert isinstance(loss, float)
|
|
assert not np.isnan(loss)
|
|
assert not np.isinf(loss)
|
|
assert loss >= 0.0 # Loss should be non-negative
|
|
|
|
self.test_results['test_training_pipeline'] = {
|
|
'status': 'PASSED',
|
|
'training_loss': float(loss),
|
|
'batch_size': len(test_predictions),
|
|
'training_successful': True
|
|
}
|
|
|
|
logger.info(f"✅ Training pipeline test PASSED (loss: {loss:.6f})")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_training_pipeline'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_trading_integration(self):
|
|
"""Test integration with trading executor"""
|
|
logger.info("💰 Testing Trading Integration...")
|
|
|
|
try:
|
|
# Initialize with simulation mode
|
|
trading_executor = TradingExecutor(simulation_mode=True)
|
|
|
|
# Test signal execution
|
|
success = trading_executor.execute_signal(
|
|
symbol='BTC/USDT',
|
|
action='BUY',
|
|
confidence=0.8,
|
|
current_price=50000.0
|
|
)
|
|
|
|
# In simulation mode, this should always succeed
|
|
assert success == True
|
|
|
|
# Check positions
|
|
positions = trading_executor.get_positions()
|
|
assert 'BTC/USDT' in positions
|
|
|
|
# Test sell signal
|
|
success = trading_executor.execute_signal(
|
|
symbol='BTC/USDT',
|
|
action='SELL',
|
|
confidence=0.8,
|
|
current_price=50100.0
|
|
)
|
|
|
|
assert success == True
|
|
|
|
# Check trade history
|
|
trade_history = trading_executor.get_trade_history()
|
|
assert len(trade_history) > 0
|
|
|
|
last_trade = trade_history[-1]
|
|
assert last_trade.symbol == 'BTC/USDT'
|
|
assert last_trade.pnl != 0 # Should have some P&L
|
|
|
|
self.test_results['test_trading_integration'] = {
|
|
'status': 'PASSED',
|
|
'simulation_mode': True,
|
|
'trades_executed': len(trade_history),
|
|
'last_trade_pnl': float(last_trade.pnl)
|
|
}
|
|
|
|
logger.info("✅ Trading integration test PASSED")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_trading_integration'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def test_performance_monitoring(self):
|
|
"""Test performance monitoring and statistics"""
|
|
logger.info("📊 Testing Performance Monitoring...")
|
|
|
|
try:
|
|
if not self.trader:
|
|
self.trading_executor = TradingExecutor(simulation_mode=True)
|
|
self.trader = RealtimeRLCOBTrader(
|
|
symbols=['BTC/USDT', 'ETH/USDT'],
|
|
trading_executor=self.trading_executor
|
|
)
|
|
|
|
# Get performance stats
|
|
stats = self.trader.get_performance_stats()
|
|
|
|
# Verify structure
|
|
assert 'symbols' in stats
|
|
assert 'training_stats' in stats
|
|
assert 'inference_stats' in stats
|
|
assert 'signal_stats' in stats
|
|
assert 'model_info' in stats
|
|
|
|
# Check symbols
|
|
assert 'BTC/USDT' in stats['symbols']
|
|
assert 'ETH/USDT' in stats['symbols']
|
|
|
|
# Check model info
|
|
for symbol in stats['symbols']:
|
|
assert symbol in stats['model_info']
|
|
model_info = stats['model_info'][symbol]
|
|
assert 'total_parameters' in model_info
|
|
assert 'trainable_parameters' in model_info
|
|
assert model_info['total_parameters'] > 0
|
|
|
|
self.test_results['test_performance_monitoring'] = {
|
|
'status': 'PASSED',
|
|
'stats_structure_valid': True,
|
|
'symbols_tracked': len(stats['symbols']),
|
|
'model_info_available': len(stats['model_info'])
|
|
}
|
|
|
|
logger.info("✅ Performance monitoring test PASSED")
|
|
|
|
except Exception as e:
|
|
self.test_results['test_performance_monitoring'] = {'status': 'FAILED', 'error': str(e)}
|
|
raise
|
|
|
|
async def generate_test_report(self):
|
|
"""Generate comprehensive test report"""
|
|
logger.info("=" * 60)
|
|
logger.info("REAL-TIME RL COB TRADER TEST REPORT")
|
|
logger.info("=" * 60)
|
|
|
|
total_tests = len(self.test_results)
|
|
passed_tests = sum(1 for result in self.test_results.values() if result['status'] == 'PASSED')
|
|
failed_tests = sum(1 for result in self.test_results.values() if result['status'] == 'FAILED')
|
|
warning_tests = sum(1 for result in self.test_results.values() if result['status'] == 'WARNING')
|
|
|
|
logger.info(f"📊 Test Summary:")
|
|
logger.info(f" Total Tests: {total_tests}")
|
|
logger.info(f" ✅ Passed: {passed_tests}")
|
|
logger.info(f" ⚠️ Warnings: {warning_tests}")
|
|
logger.info(f" ❌ Failed: {failed_tests}")
|
|
|
|
success_rate = (passed_tests / total_tests) * 100 if total_tests > 0 else 0
|
|
logger.info(f" Success Rate: {success_rate:.1f}%")
|
|
|
|
logger.info("\n📋 Detailed Results:")
|
|
for test_name, result in self.test_results.items():
|
|
status_icon = "✅" if result['status'] == 'PASSED' else "⚠️" if result['status'] == 'WARNING' else "❌"
|
|
logger.info(f" {status_icon} {test_name}: {result['status']}")
|
|
|
|
if result['status'] == 'FAILED':
|
|
logger.error(f" Error: {result.get('error', 'Unknown error')}")
|
|
|
|
# System readiness assessment
|
|
logger.info("\n🎯 System Readiness Assessment:")
|
|
if failed_tests == 0:
|
|
if warning_tests == 0:
|
|
logger.info(" 🟢 SYSTEM READY FOR DEPLOYMENT")
|
|
logger.info(" All tests passed. The real-time RL COB trader is ready for live operation.")
|
|
else:
|
|
logger.info(" 🟡 SYSTEM READY WITH WARNINGS")
|
|
logger.info(" System is functional but some performance warnings exist.")
|
|
else:
|
|
logger.info(" 🔴 SYSTEM NOT READY")
|
|
logger.info(" Critical issues found. Fix errors before deployment.")
|
|
|
|
# Save detailed report
|
|
report_data = {
|
|
'timestamp': datetime.now().isoformat(),
|
|
'test_summary': {
|
|
'total_tests': total_tests,
|
|
'passed_tests': passed_tests,
|
|
'warning_tests': warning_tests,
|
|
'failed_tests': failed_tests,
|
|
'success_rate': success_rate
|
|
},
|
|
'test_results': self.test_results,
|
|
'system_readiness': 'READY' if failed_tests == 0 else 'NOT_READY'
|
|
}
|
|
|
|
report_file = f"test_reports/realtime_rl_test_report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json"
|
|
|
|
import os
|
|
os.makedirs('test_reports', exist_ok=True)
|
|
|
|
with open(report_file, 'w') as f:
|
|
json.dump(report_data, f, indent=2, default=str)
|
|
|
|
logger.info(f"\n📄 Detailed report saved to: {report_file}")
|
|
logger.info("=" * 60)
|
|
|
|
async def main():
|
|
"""Main test entry point"""
|
|
logger.info("Starting Real-time RL COB Trader Test Suite...")
|
|
|
|
tester = RealtimeRLTester()
|
|
await tester.run_all_tests()
|
|
|
|
if __name__ == "__main__":
|
|
# Set event loop policy for Windows compatibility
|
|
if hasattr(asyncio, 'WindowsProactorEventLoopPolicy'):
|
|
asyncio.set_event_loop_policy(asyncio.WindowsProactorEventLoopPolicy())
|
|
|
|
asyncio.run(main()) |