restart script

NN/models/cnn_model.py

@@ -69,20 +69,30 @@ class ResidualBlock(nn.Module):
         super().__init__()
         self.conv1 = nn.Conv1d(channels, channels, kernel_size=3, padding=1)
         self.conv2 = nn.Conv1d(channels, channels, kernel_size=3, padding=1)
-        self.norm1 = nn.BatchNorm1d(channels)
+        self.norm1 = nn.GroupNorm(1, channels)  # Changed from BatchNorm1d to GroupNorm
-        self.norm2 = nn.BatchNorm1d(channels)
+        self.norm2 = nn.GroupNorm(1, channels)  # Changed from BatchNorm1d to GroupNorm
         self.dropout = nn.Dropout(dropout)
         
     def forward(self, x: torch.Tensor) -> torch.Tensor:
-        residual = x
+        # Create completely independent copy for residual connection
+        residual = x.detach().clone()
         
-        out = F.relu(self.norm1(self.conv1(x)))
+        # First convolution branch - ensure no memory sharing
+        out = self.conv1(x)
+        out = self.norm1(out)
+        out = F.relu(out)
         out = self.dropout(out)
-        out = self.norm2(self.conv2(out))
         
-        # Add residual connection (avoid in-place operation)
+        # Second convolution branch
-        out = out + residual
+        out = self.conv2(out)
-        return F.relu(out)
+        out = self.norm2(out)
+        
+        # Residual connection - create completely new tensor
+        # Avoid any potential in-place operations or memory sharing
+        combined = residual + out
+        result = F.relu(combined)
+        
+        return result
 
 class SpatialAttentionBlock(nn.Module):
     """Spatial attention for feature maps"""
@@ -144,11 +154,11 @@ class EnhancedCNNModel(nn.Module):
         # Feature fusion with more capacity
         self.feature_fusion = nn.Sequential(
             nn.Conv1d(base_channels * 4, base_channels * 3, kernel_size=1),  # 4 paths now
-            nn.BatchNorm1d(base_channels * 3),
+            nn.GroupNorm(1, base_channels * 3),  # Changed from BatchNorm1d to GroupNorm
             nn.ReLU(),
             nn.Dropout(dropout_rate),
             nn.Conv1d(base_channels * 3, base_channels * 2, kernel_size=1),
-            nn.BatchNorm1d(base_channels * 2),
+            nn.GroupNorm(1, base_channels * 2),  # Changed from BatchNorm1d to GroupNorm
             nn.ReLU(),
             nn.Dropout(dropout_rate)
         )
@@ -258,22 +268,22 @@ class EnhancedCNNModel(nn.Module):
         
         # Initialize weights
         self._initialize_weights()
-        
+    
     def _build_conv_path(self, in_channels: int, out_channels: int, kernel_size: int) -> nn.Module:
         """Build a convolutional path with multiple layers"""
         return nn.Sequential(
             nn.Conv1d(in_channels, out_channels, kernel_size, padding=kernel_size//2),
-            nn.BatchNorm1d(out_channels),
+            nn.GroupNorm(1, out_channels),  # Changed from BatchNorm1d to GroupNorm
             nn.ReLU(),
             nn.Dropout(0.1),
             
             nn.Conv1d(out_channels, out_channels, kernel_size, padding=kernel_size//2),
-            nn.BatchNorm1d(out_channels),
+            nn.GroupNorm(1, out_channels),  # Changed from BatchNorm1d to GroupNorm
             nn.ReLU(),
             nn.Dropout(0.1),
             
             nn.Conv1d(out_channels, out_channels, kernel_size, padding=kernel_size//2),
-            nn.BatchNorm1d(out_channels),
+            nn.GroupNorm(1, out_channels),  # Changed from BatchNorm1d to GroupNorm
             nn.ReLU()
         )
     
@@ -288,19 +298,28 @@ class EnhancedCNNModel(nn.Module):
                 nn.init.xavier_normal_(m.weight)
                 if m.bias is not None:
                     nn.init.constant_(m.bias, 0)
-            elif isinstance(m, nn.BatchNorm1d):
+            elif isinstance(m, (nn.BatchNorm1d, nn.GroupNorm, nn.LayerNorm)):
-                nn.init.constant_(m.weight, 1)
+                if hasattr(m, 'weight') and m.weight is not None:
-                nn.init.constant_(m.bias, 0)
+                    nn.init.constant_(m.weight, 1)
+                if hasattr(m, 'bias') and m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+    
+    def _memory_barrier(self, tensor: torch.Tensor) -> torch.Tensor:
+        """Create a memory barrier to prevent in-place operation issues"""
+        return tensor.detach().clone().requires_grad_(tensor.requires_grad)
     
     def forward(self, x: torch.Tensor) -> Dict[str, torch.Tensor]:
         """
-        Forward pass with multiple outputs
+        Forward pass with multiple outputs - completely avoiding in-place operations
         Args:
             x: Input tensor of shape [batch_size, sequence_length, features]
         Returns:
             Dictionary with predictions, confidence, regime, and volatility
         """
-        # Handle input shapes flexibly
+        # Apply memory barrier to input
+        x = self._memory_barrier(x)
+        
+        # Handle input shapes flexibly - create new tensors to avoid memory sharing
         if len(x.shape) == 2:
             # Input is [seq_len, features] - add batch dimension
             x = x.unsqueeze(0)
@@ -308,76 +327,96 @@ class EnhancedCNNModel(nn.Module):
             # Input has extra dimensions - flatten to [batch, seq, features]
             x = x.view(x.shape[0], -1, x.shape[-1])
         
+        x = self._memory_barrier(x)  # Apply barrier after shape changes
         batch_size, seq_len, features = x.shape
         
         # Reshape for processing: [batch, seq, features] -> [batch*seq, features]
         x_reshaped = x.view(-1, features)
+        x_reshaped = self._memory_barrier(x_reshaped)
         
         # Input embedding
         embedded = self.input_embedding(x_reshaped)  # [batch*seq, base_channels]
+        embedded = self._memory_barrier(embedded)
         
         # Reshape back for conv1d: [batch*seq, channels] -> [batch, channels, seq]
-        embedded = embedded.view(batch_size, seq_len, -1).transpose(1, 2)
+        embedded = embedded.view(batch_size, seq_len, -1).transpose(1, 2).contiguous()
+        embedded = self._memory_barrier(embedded)
         
-        # Multi-scale feature extraction
+        # Multi-scale feature extraction - ensure each path creates independent tensors
-        path1 = self.conv_path1(embedded)
+        path1 = self._memory_barrier(self.conv_path1(embedded))
-        path2 = self.conv_path2(embedded)
+        path2 = self._memory_barrier(self.conv_path2(embedded))
-        path3 = self.conv_path3(embedded)
+        path3 = self._memory_barrier(self.conv_path3(embedded))
-        path4 = self.conv_path4(embedded)
+        path4 = self._memory_barrier(self.conv_path4(embedded))
         
-        # Feature fusion
+        # Feature fusion - create new tensor
         fused_features = torch.cat([path1, path2, path3, path4], dim=1)
-        fused_features = self.feature_fusion(fused_features)
+        fused_features = self._memory_barrier(self.feature_fusion(fused_features))
         
         # Apply residual blocks with spatial attention
-        current_features = fused_features
+        current_features = self._memory_barrier(fused_features)
         for i, (res_block, attention) in enumerate(zip(self.residual_blocks, self.spatial_attention)):
-            current_features = res_block(current_features)
+            current_features = self._memory_barrier(res_block(current_features))
             if i % 2 == 0:  # Apply attention every other block
-                current_features = attention(current_features)
+                current_features = self._memory_barrier(attention(current_features))
         
         # Apply remaining residual blocks
         for res_block in self.residual_blocks[len(self.spatial_attention):]:
-            current_features = res_block(current_features)
+            current_features = self._memory_barrier(res_block(current_features))
         
         # Temporal attention - apply both attention layers
         # Reshape for attention: [batch, channels, seq] -> [batch, seq, channels]
-        attention_input = current_features.transpose(1, 2)
+        attention_input = current_features.transpose(1, 2).contiguous()
-        attended_features = self.temporal_attention1(attention_input)
+        attention_input = self._memory_barrier(attention_input)
-        attended_features = self.temporal_attention2(attended_features)
+        
+        attended_features = self._memory_barrier(self.temporal_attention1(attention_input))
+        attended_features = self._memory_barrier(self.temporal_attention2(attended_features))
         # Back to conv format: [batch, seq, channels] -> [batch, channels, seq]
-        attended_features = attended_features.transpose(1, 2)
+        attended_features = attended_features.transpose(1, 2).contiguous()
+        attended_features = self._memory_barrier(attended_features)
         
-        # Global aggregation
+        # Global aggregation - create independent tensors
-        avg_pooled = self.global_pool(attended_features).squeeze(-1)  # [batch, channels]
+        avg_pooled = self.global_pool(attended_features)
-        max_pooled = self.global_max_pool(attended_features).squeeze(-1)  # [batch, channels]
+        avg_pooled = self._memory_barrier(avg_pooled.view(avg_pooled.shape[0], -1))  # Flatten instead of squeeze
         
-        # Combine global features
+        max_pooled = self.global_max_pool(attended_features) 
+        max_pooled = self._memory_barrier(max_pooled.view(max_pooled.shape[0], -1))  # Flatten instead of squeeze
+        
+        # Combine global features - create new tensor
         global_features = torch.cat([avg_pooled, max_pooled], dim=1)
+        global_features = self._memory_barrier(global_features)
         
         # Advanced feature processing
-        processed_features = self.advanced_features(global_features)
+        processed_features = self._memory_barrier(self.advanced_features(global_features))
         
-        # Multi-task predictions
+        # Multi-task predictions - ensure each creates independent tensors
-        regime_probs = self.regime_detector(processed_features)
+        regime_probs = self._memory_barrier(self.regime_detector(processed_features))
-        volatility_pred = self.volatility_predictor(processed_features)
+        volatility_pred = self._memory_barrier(self.volatility_predictor(processed_features))
-        confidence = self.confidence_head(processed_features)
+        confidence = self._memory_barrier(self.confidence_head(processed_features))
         
         # Combine all features for final decision (8 regime classes + 1 volatility)
-        combined_features = torch.cat([processed_features, regime_probs, volatility_pred], dim=1)
+        # Create completely independent tensors for concatenation
-        trading_logits = self.decision_head(combined_features)
+        vol_pred_flat = self._memory_barrier(volatility_pred.view(volatility_pred.shape[0], -1))  # Flatten instead of squeeze
+        combined_features = torch.cat([processed_features, regime_probs, vol_pred_flat], dim=1)
+        combined_features = self._memory_barrier(combined_features)
         
-        # Apply temperature scaling for better calibration
+        trading_logits = self._memory_barrier(self.decision_head(combined_features))
+        
+        # Apply temperature scaling for better calibration - create new tensor
         temperature = 1.5
-        trading_probs = F.softmax(trading_logits / temperature, dim=1)
+        scaled_logits = trading_logits / temperature
+        trading_probs = self._memory_barrier(F.softmax(scaled_logits, dim=1))
+        
+        # Flatten confidence to ensure consistent shape
+        confidence_flat = self._memory_barrier(confidence.view(confidence.shape[0], -1))
+        volatility_flat = self._memory_barrier(volatility_pred.view(volatility_pred.shape[0], -1))
         
         return {
-            'logits': trading_logits,
+            'logits': self._memory_barrier(trading_logits),
-            'probabilities': trading_probs,
+            'probabilities': self._memory_barrier(trading_probs),
-            'confidence': confidence.squeeze(-1),
+            'confidence': confidence_flat[:, 0] if confidence_flat.shape[1] > 0 else confidence_flat.view(-1)[0],
-            'regime': regime_probs,
+            'regime': self._memory_barrier(regime_probs),
-            'volatility': volatility_pred.squeeze(-1),
+            'volatility': volatility_flat[:, 0] if volatility_flat.shape[1] > 0 else volatility_flat.view(-1)[0],
-            'features': processed_features
+            'features': self._memory_barrier(processed_features)
         }
     
     def predict(self, feature_matrix: np.ndarray) -> Dict[str, Any]:
@@ -478,60 +517,128 @@ class CNNModelTrainer:
         
         self.training_history = []
         
+    def reset_computational_graph(self):
+        """Reset the computational graph to prevent in-place operation issues"""
+        try:
+            # Clear all gradients
+            for param in self.model.parameters():
+                param.grad = None
+            
+            # Force garbage collection
+            import gc
+            gc.collect()
+            
+            # Clear CUDA cache if available
+            if torch.cuda.is_available():
+                torch.cuda.empty_cache()
+                torch.cuda.synchronize()
+            
+            # Reset optimizer state if needed
+            for group in self.optimizer.param_groups:
+                for param in group['params']:
+                    if param in self.optimizer.state:
+                        # Clear momentum buffers that might have stale references
+                        self.optimizer.state[param] = {}
+                        
+        except Exception as e:
+            logger.warning(f"Error during computational graph reset: {e}")
+    
     def train_step(self, x: torch.Tensor, y: torch.Tensor, 
                    confidence_targets: Optional[torch.Tensor] = None,
                    regime_targets: Optional[torch.Tensor] = None,
                    volatility_targets: Optional[torch.Tensor] = None) -> Dict[str, float]:
-        """Single training step with multi-task learning"""
+        """Single training step with multi-task learning and robust error handling"""
         
-        self.model.train()
+        # Reset computational graph before each training step
-        self.optimizer.zero_grad()
+        self.reset_computational_graph()
         
-        # Forward pass
+        try:
-        outputs = self.model(x)
+            self.model.train()
-        
+            
-        # Main trading loss
+            # Ensure inputs are completely independent from original tensors
-        main_loss = self.main_criterion(outputs['logits'], y)
+            x_train = x.detach().clone().requires_grad_(False).to(self.device)
-        total_loss = main_loss
+            y_train = y.detach().clone().requires_grad_(False).to(self.device)
-        
+            
-        losses = {'main_loss': main_loss.item()}
+            # Forward pass with error handling
-        
+            try:
-        # Confidence loss (if targets provided)
+                outputs = self.model(x_train)
-        if confidence_targets is not None:
+            except RuntimeError as forward_error:
-            conf_loss = self.confidence_criterion(outputs['confidence'], confidence_targets)
+                if "modified by an inplace operation" in str(forward_error):
-            total_loss += 0.1 * conf_loss
+                    logger.error(f"In-place operation in forward pass: {forward_error}")
-            losses['confidence_loss'] = conf_loss.item()
+                    self.reset_computational_graph()
-        
+                    return {'main_loss': 0.0, 'total_loss': 0.0, 'accuracy': 0.5}
-        # Regime classification loss (if targets provided)
+                else:
-        if regime_targets is not None:
+                    raise forward_error
-            regime_loss = self.regime_criterion(outputs['regime'], regime_targets)
+            
-            total_loss += 0.05 * regime_loss
+            # Calculate main loss with detached outputs to prevent memory sharing
-            losses['regime_loss'] = regime_loss.item()
+            main_loss = self.main_criterion(outputs['logits'], y_train)
-        
+            total_loss = main_loss
-        # Volatility prediction loss (if targets provided)
+            
-        if volatility_targets is not None:
+            losses = {'main_loss': main_loss.item()}
-            vol_loss = self.volatility_criterion(outputs['volatility'], volatility_targets)
+            
-            total_loss += 0.05 * vol_loss
+            # Add auxiliary losses if targets provided
-            losses['volatility_loss'] = vol_loss.item()
+            if confidence_targets is not None:
-        
+                conf_targets = confidence_targets.detach().clone().to(self.device)
-        losses['total_loss'] = total_loss.item()
+                conf_loss = self.confidence_criterion(outputs['confidence'], conf_targets)
-        
+                total_loss = total_loss + 0.1 * conf_loss
-        # Backward pass
+                losses['confidence_loss'] = conf_loss.item()
-        total_loss.backward()
+            
-        
+            if regime_targets is not None:
-        # Gradient clipping
+                regime_targets_clean = regime_targets.detach().clone().to(self.device)
-        torch.nn.utils.clip_grad_norm_(self.model.parameters(), max_norm=1.0)
+                regime_loss = self.regime_criterion(outputs['regime'], regime_targets_clean)
-        
+                total_loss = total_loss + 0.05 * regime_loss
-        self.optimizer.step()
+                losses['regime_loss'] = regime_loss.item()
-        self.scheduler.step()
+            
-        
+            if volatility_targets is not None:
-        # Calculate accuracy
+                vol_targets = volatility_targets.detach().clone().to(self.device)
-        with torch.no_grad():
+                vol_loss = self.volatility_criterion(outputs['volatility'], vol_targets)
-            predictions = torch.argmax(outputs['probabilities'], dim=1)
+                total_loss = total_loss + 0.05 * vol_loss
-            accuracy = (predictions == y).float().mean().item()
+                losses['volatility_loss'] = vol_loss.item()
-            losses['accuracy'] = accuracy
+            
-        
+            losses['total_loss'] = total_loss.item()
-        return losses
+            
+            # Backward pass with comprehensive error handling
+            try:
+                total_loss.backward()
+                
+            except RuntimeError as backward_error:
+                if "modified by an inplace operation" in str(backward_error):
+                    logger.error(f"In-place operation during backward pass: {backward_error}")
+                    logger.error("Attempting to continue training with gradient reset...")
+                    
+                    # Comprehensive cleanup
+                    self.reset_computational_graph()
+                    
+                    return {'main_loss': losses.get('main_loss', 0.0), 'total_loss': losses.get('total_loss', 0.0), 'accuracy': 0.5}
+                else:
+                    raise backward_error
+            
+            # Gradient clipping
+            torch.nn.utils.clip_grad_norm_(self.model.parameters(), max_norm=1.0)
+            
+            # Optimizer step
+            self.optimizer.step()
+            self.scheduler.step()
+            
+            # Calculate accuracy with detached tensors
+            with torch.no_grad():
+                predictions = torch.argmax(outputs['probabilities'], dim=1)
+                accuracy = (predictions == y_train).float().mean().item()
+                losses['accuracy'] = accuracy
+            
+            return losses
+            
+        except Exception as e:
+            logger.error(f"Training step failed with unexpected error: {e}")
+            logger.error(f"Error type: {type(e).__name__}")
+            import traceback
+            logger.error(f"Full traceback: {traceback.format_exc()}")
+            
+            # Comprehensive cleanup on any error
+            self.reset_computational_graph()
+            
+            # Return safe dummy values to continue training
+            return {'main_loss': 0.0, 'total_loss': 0.0, 'accuracy': 0.5}
     
     def save_model(self, filepath: str, metadata: Optional[Dict] = None):
         """Save model with metadata"""
@@ -610,7 +717,7 @@ class CNNModel:
             feature_dim=input_shape[1],
             output_size=output_size
         )
-        self.trainer = CNNModelTrainer(self.model, device=self.device)
+        self.trainer = CNNModelTrainer(self.model, device=str(self.device))
         
         logger.info(f"CNN Model wrapper initialized: input_shape={input_shape}, output_size={output_size}")
         

restart_main_overnight.ps1

@@ -0,0 +1,90 @@
+# Overnight Training Restart Script (PowerShell)
+# Keeps main.py running continuously, restarting it if it crashes.
+# Usage: .\restart_main_overnight.ps1
+
+Write-Host "=" * 60
+Write-Host "OVERNIGHT TRAINING RESTART SCRIPT (PowerShell)"
+Write-Host "=" * 60
+Write-Host "Press Ctrl+C to stop the restart loop"
+Write-Host "Main script: main.py" 
+Write-Host "Restart delay on crash: 10 seconds"
+Write-Host "=" * 60
+
+$restartCount = 0
+$startTime = Get-Date
+
+# Create logs directory if it doesn't exist
+if (!(Test-Path "logs")) {
+    New-Item -ItemType Directory -Path "logs"
+}
+
+# Setup log file
+$timestamp = Get-Date -Format "yyyyMMdd_HHmmss"
+$logFile = "logs\restart_main_ps_$timestamp.log"
+
+function Write-Log {
+    param($Message)
+    $timestamp = Get-Date -Format "yyyy-MM-dd HH:mm:ss"
+    $logMessage = "$timestamp - $Message"
+    Write-Host $logMessage
+    Add-Content -Path $logFile -Value $logMessage
+}
+
+Write-Log "Restart script started, logging to: $logFile"
+
+# Kill any existing Python processes
+try {
+    Get-Process python* -ErrorAction SilentlyContinue | Stop-Process -Force -ErrorAction SilentlyContinue
+    Start-Sleep -Seconds 2
+    Write-Log "Killed existing Python processes"
+} catch {
+    Write-Log "Could not kill existing processes: $_"
+}
+
+try {
+    while ($true) {
+        $restartCount++
+        $runStartTime = Get-Date
+        
+        Write-Log "[RESTART #$restartCount] Starting main.py at $(Get-Date -Format 'HH:mm:ss')"
+        
+        # Start main.py
+        try {
+            $process = Start-Process -FilePath "python" -ArgumentList "main.py" -PassThru -Wait
+            $exitCode = $process.ExitCode
+            $runEndTime = Get-Date
+            $runDuration = ($runEndTime - $runStartTime).TotalSeconds
+            
+            Write-Log "[EXIT] main.py exited with code $exitCode"
+            Write-Log "[DURATION] Process ran for $([math]::Round($runDuration, 1)) seconds"
+            
+            # Check for fast exits
+            if ($runDuration -lt 30) {
+                Write-Log "[FAST EXIT] Process exited quickly, waiting 30 seconds..."
+                Start-Sleep -Seconds 30
+            } else {
+                Write-Log "[DELAY] Waiting 10 seconds before restart..."
+                Start-Sleep -Seconds 10
+            }
+            
+            # Log stats every 10 restarts
+            if ($restartCount % 10 -eq 0) {
+                $totalDuration = (Get-Date) - $startTime
+                Write-Log "[STATS] Session: $restartCount restarts in $([math]::Round($totalDuration.TotalHours, 1)) hours"
+            }
+            
+        } catch {
+            Write-Log "[ERROR] Error starting main.py: $_"
+            Start-Sleep -Seconds 10
+        }
+    }
+} catch {
+    Write-Log "[INTERRUPT] Restart loop interrupted: $_"
+} finally {
+    $totalDuration = (Get-Date) - $startTime
+    Write-Log "=" * 60
+    Write-Log "OVERNIGHT TRAINING SESSION COMPLETE"
+    Write-Log "Total restarts: $restartCount"
+    Write-Log "Total session time: $([math]::Round($totalDuration.TotalHours, 1)) hours"
+    Write-Log "=" * 60
+} 

restart_main_overnight.py

@@ -0,0 +1,188 @@
+#!/usr/bin/env python3
+"""
+Overnight Training Restart Script
+Keeps main.py running continuously, restarting it if it crashes.
+Designed for overnight training sessions with unstable code.
+
+Usage:
+    python restart_main_overnight.py
+
+Press Ctrl+C to stop the restart loop.
+"""
+
+import subprocess
+import sys
+import time
+import logging
+from datetime import datetime
+from pathlib import Path
+import signal
+import os
+
+# Setup logging for the restart script
+def setup_restart_logging():
+    """Setup logging for restart events"""
+    log_dir = Path("logs")
+    log_dir.mkdir(exist_ok=True)
+    
+    # Create restart log file with timestamp
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    log_file = log_dir / f"restart_main_{timestamp}.log"
+    
+    logging.basicConfig(
+        level=logging.INFO,
+        format='%(asctime)s - %(levelname)s - %(message)s',
+        handlers=[
+            logging.FileHandler(log_file, encoding='utf-8'),
+            logging.StreamHandler(sys.stdout)
+        ]
+    )
+    
+    logger = logging.getLogger(__name__)
+    logger.info(f"Restart script logging to: {log_file}")
+    return logger
+
+def kill_existing_processes(logger):
+    """Kill any existing main.py processes to avoid conflicts"""
+    try:
+        if os.name == 'nt':  # Windows
+            # Kill any existing Python processes running main.py
+            subprocess.run(['taskkill', '/f', '/im', 'python.exe'], 
+                         capture_output=True, check=False)
+            subprocess.run(['taskkill', '/f', '/im', 'pythonw.exe'], 
+                         capture_output=True, check=False)
+            time.sleep(2)
+    except Exception as e:
+        logger.warning(f"Could not kill existing processes: {e}")
+
+def run_main_with_restart(logger):
+    """Main restart loop"""
+    restart_count = 0
+    consecutive_fast_exits = 0
+    start_time = datetime.now()
+    
+    logger.info("=" * 60)
+    logger.info("OVERNIGHT TRAINING RESTART SCRIPT STARTED")
+    logger.info("=" * 60)
+    logger.info("Press Ctrl+C to stop the restart loop")
+    logger.info("Main script: main.py")
+    logger.info("Restart delay on crash: 10 seconds")
+    logger.info("Fast exit protection: Enabled")
+    logger.info("=" * 60)
+    
+    # Kill any existing processes
+    kill_existing_processes(logger)
+    
+    while True:
+        try:
+            restart_count += 1
+            run_start_time = datetime.now()
+            
+            logger.info(f"[RESTART #{restart_count}] Starting main.py at {run_start_time.strftime('%H:%M:%S')}")
+            
+            # Start main.py as subprocess
+            process = subprocess.Popen([
+                sys.executable, "main.py"
+            ], stdout=subprocess.PIPE, stderr=subprocess.STDOUT, 
+               universal_newlines=True, bufsize=1)
+            
+            logger.info(f"[PROCESS] main.py started with PID: {process.pid}")
+            
+            # Stream output from main.py
+            try:
+                if process.stdout:
+                    while True:
+                        output = process.stdout.readline()
+                        if output == '' and process.poll() is not None:
+                            break
+                        if output:
+                            # Forward output from main.py (remove extra newlines)
+                            print(f"[MAIN] {output.rstrip()}")
+                else:
+                    # If no stdout, just wait for process to complete
+                    process.wait()
+            except KeyboardInterrupt:
+                logger.info("[INTERRUPT] Ctrl+C received, stopping main.py...")
+                process.terminate()
+                try:
+                    process.wait(timeout=10)
+                except subprocess.TimeoutExpired:
+                    logger.warning("[FORCE KILL] Process didn't terminate, force killing...")
+                    process.kill()
+                raise
+            
+            # Process has exited
+            exit_code = process.poll()
+            run_end_time = datetime.now()
+            run_duration = (run_end_time - run_start_time).total_seconds()
+            
+            logger.info(f"[EXIT] main.py exited with code {exit_code}")
+            logger.info(f"[DURATION] Process ran for {run_duration:.1f} seconds")
+            
+            # Check for fast exits (potential configuration issues)
+            if run_duration < 30:  # Less than 30 seconds
+                consecutive_fast_exits += 1
+                logger.warning(f"[FAST EXIT] Process exited quickly ({consecutive_fast_exits} consecutive)")
+                
+                if consecutive_fast_exits >= 5:
+                    logger.error("[ABORT] Too many consecutive fast exits (5+)")
+                    logger.error("This indicates a configuration or startup problem")
+                    logger.error("Please check the main.py script manually")
+                    break
+                    
+                # Longer delay for fast exits
+                delay = min(60, 10 * consecutive_fast_exits)
+                logger.info(f"[DELAY] Waiting {delay} seconds before restart due to fast exit...")
+                time.sleep(delay)
+            else:
+                consecutive_fast_exits = 0  # Reset counter
+                logger.info("[DELAY] Waiting 10 seconds before restart...")
+                time.sleep(10)
+            
+            # Log session statistics every 10 restarts
+            if restart_count % 10 == 0:
+                total_duration = (datetime.now() - start_time).total_seconds()
+                logger.info(f"[STATS] Session: {restart_count} restarts in {total_duration/3600:.1f} hours")
+                
+        except KeyboardInterrupt:
+            logger.info("[SHUTDOWN] Restart loop interrupted by user")
+            break
+        except Exception as e:
+            logger.error(f"[ERROR] Unexpected error in restart loop: {e}")
+            logger.error("Continuing restart loop after 30 second delay...")
+            time.sleep(30)
+    
+    total_duration = (datetime.now() - start_time).total_seconds()
+    logger.info("=" * 60)
+    logger.info("OVERNIGHT TRAINING SESSION COMPLETE")
+    logger.info(f"Total restarts: {restart_count}")
+    logger.info(f"Total session time: {total_duration/3600:.1f} hours")
+    logger.info("=" * 60)
+
+def main():
+    """Main entry point"""
+    # Setup signal handlers for clean shutdown
+    def signal_handler(signum, frame):
+        logger.info(f"[SIGNAL] Received signal {signum}, shutting down...")
+        sys.exit(0)
+    
+    signal.signal(signal.SIGINT, signal_handler)
+    if hasattr(signal, 'SIGTERM'):
+        signal.signal(signal.SIGTERM, signal_handler)
+    
+    # Setup logging
+    global logger
+    logger = setup_restart_logging()
+    
+    try:
+        run_main_with_restart(logger)
+    except Exception as e:
+        logger.error(f"[FATAL] Fatal error in restart script: {e}")
+        import traceback
+        logger.error(traceback.format_exc())
+        return 1
+    
+    return 0
+
+if __name__ == "__main__":
+    sys.exit(main())