scripts/training/final_breakthrough_training.py

#!/usr/bin/env python3
"""
Final Breakthrough BitTransformerLM Training Script
=================================================

The complete training script using the ACTUAL BitTransformerLM model
with the breakthrough Fixed RL Adafactor configuration and full
HuggingFace dataset support with checkpoint resumption.
"""

import sys
import os
import json
import logging
from pathlib import Path
from datetime import datetime
from typing import Optional, Dict, Any

import torch
import torch.nn.functional as F
from datasets import load_dataset
from huggingface_hub import login

# Add paths for imports
sys.path.append('/data')
sys.path.append('/data/BitTransformerLM')

from bit_transformer import BitTransformerLM, text_to_bits
from BTLM_Extensions import configure_adafactor_optimizer

# Setup logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler('/data/BitTransformerLM/breakthrough_training.log'),
        logging.StreamHandler()
    ]
)
logger = logging.getLogger(__name__)

class BreakthroughTrainer:
    """Production-grade BitTransformerLM trainer with breakthrough configuration."""
    
    def __init__(self, config: Dict[str, Any]):
        self.config = config
        self.device = torch.device('cpu')  # CPU training as per breakthrough
        self.model = None
        self.optimizer = None
        self.scheduler = None
        self.dataset = None
        self.checkpoint_dir = Path(config['checkpoint_dir'])
        self.checkpoint_dir.mkdir(parents=True, exist_ok=True)
        
        # Training state
        self.current_epoch = 0
        self.total_steps = 0
        self.best_loss = float('inf')
        self.training_history = []

    def load_and_prepare_dataset(self):
        """Load HF dataset and convert to proper bit tensors."""
        logger.info("Loading WCNegentropy/BitTransformerLM dataset...")
        
        # Login to HuggingFace
        login(token=self.config['hf_token'])
        
        # Load dataset
        dataset = load_dataset("WCNegentropy/BitTransformerLM")
        train_data = dataset['train']
        
        logger.info(f"Dataset loaded: {len(train_data)} samples")
        
        # Process dataset - convert to bits using the ACTUAL text_to_bits function
        bit_sequences = []
        for i, sample in enumerate(train_data):
            if i % 1000 == 0:
                logger.info(f"Processing sample {i}/{len(train_data)}")
            
            # Try to get text from various fields
            text = None
            if 'original_text' in sample and sample['original_text']:
                text = sample['original_text']
            elif 'text' in sample and sample['text']:
                text = sample['text']
            
            if text and text.strip():
                # Use ACTUAL text_to_bits function
                bits = text_to_bits(text)
                if len(bits) >= self.config['sequence_length']:
                    bit_sequences.append(bits)
        
        logger.info(f"Processed {len(bit_sequences)} valid bit sequences")
        
        # Create training sequences with proper length
        seq_len = self.config['sequence_length']
        training_sequences = []
        
        for bits in bit_sequences:
            # Create overlapping chunks
            for i in range(0, len(bits) - seq_len + 1, seq_len // 2):
                chunk = bits[i:i + seq_len]
                if len(chunk) == seq_len:
                    training_sequences.append(chunk)
        
        # Convert to tensor with proper dtype
        self.dataset = torch.tensor(training_sequences, dtype=torch.long)
        logger.info(f"Created training dataset: {self.dataset.shape}")
        
        return self.dataset

    def create_breakthrough_model(self):
        """Create the EXACT breakthrough 16M parameter BitTransformerLM."""
        logger.info("Creating breakthrough 16M parameter BitTransformerLM...")
        
        # BREAKTHROUGH CONFIGURATION - exactly as identified before
        self.model = BitTransformerLM(
            d_model=512,            # Breakthrough config
            nhead=16,              # 16 attention heads  
            num_layers=8,          # 8 layers for ~16M params
            dim_feedforward=1024,  # 2x d_model
            max_seq_len=self.config['sequence_length'],
            lambda_K=0.05,         # Safety telemetry weights
            lambda_C=0.05,
            lambda_S=0.05,
            reversible=True,       # Memory efficiency
            use_checkpoint=True,   # Gradient checkpointing
            use_autocast=True,     # CPU mixed precision
            use_act=True,          # Adaptive Computation Time
            act_threshold=0.9
        ).to(self.device)
        
        # Calculate and verify parameter count
        total_params = sum(p.numel() for p in self.model.parameters())
        trainable_params = sum(p.numel() for p in self.model.parameters() if p.requires_grad)
        
        logger.info(f"Model created: {total_params:,} total parameters ({trainable_params:,} trainable)")
        logger.info(f"Target: ~16M parameters - {'✓' if 15_000_000 <= total_params <= 17_000_000 else '✗'}")
        
        return self.model

    def setup_optimizer(self):
        """Setup Fixed RL Adafactor optimizer (the breakthrough secret sauce)."""
        logger.info("Setting up Fixed RL Adafactor optimizer...")
        
        # Calculate total steps
        steps_per_epoch = len(self.dataset) // self.config['batch_size']
        total_steps = steps_per_epoch * self.config['num_epochs']
        
        # CRITICAL: Use FIXED LR, not auto-LR (the breakthrough discovery!)
        self.optimizer, self.scheduler = configure_adafactor_optimizer(
            self.model,
            lr=self.config['learning_rate'],  # FIXED LR - key to breakthrough!
            weight_decay=self.config['weight_decay'],
            total_steps=total_steps
        )
        
        logger.info(f"Fixed RL Adafactor configured with LR={self.config['learning_rate']}")
        logger.info(f"Total training steps: {total_steps}")
        
        return self.optimizer, self.scheduler

    def save_checkpoint(self, epoch: int, loss: float, is_best: bool = False):
        """Save complete model checkpoint with all training state."""
        checkpoint_data = {
            'epoch': epoch,
            'total_steps': self.total_steps,
            'model_state_dict': self.model.state_dict(),
            'optimizer_state_dict': self.optimizer.state_dict(),
            'scheduler_state_dict': self.scheduler.state_dict() if self.scheduler else None,
            'loss': loss,
            'best_loss': self.best_loss,
            'config': self.config,
            'training_history': self.training_history,
            'timestamp': datetime.now().isoformat(),
            'model_config': self.model._current_params()  # Save model hyperparams
        }
        
        # Save latest checkpoint
        latest_path = self.checkpoint_dir / 'checkpoint_latest.pt'
        torch.save(checkpoint_data, latest_path)
        logger.info(f"Saved checkpoint: {latest_path}")
        
        # Save epoch-specific checkpoint
        epoch_path = self.checkpoint_dir / f'checkpoint_epoch_{epoch:04d}.pt'
        torch.save(checkpoint_data, epoch_path)
        
        # Save best model if this is the best loss
        if is_best:
            best_path = self.checkpoint_dir / 'checkpoint_best.pt'
            torch.save(checkpoint_data, best_path)
            logger.info(f"🏆 NEW BEST MODEL! Loss: {loss:.6f} -> {best_path}")
        
        # Save training config for reference
        config_path = self.checkpoint_dir / 'training_config.json'
        with open(config_path, 'w') as f:
            json.dump(self.config, f, indent=2)

    def load_checkpoint(self, checkpoint_path: Optional[str] = None) -> bool:
        """Load checkpoint if available and resume training."""
        if checkpoint_path is None:
            checkpoint_path = self.checkpoint_dir / 'checkpoint_latest.pt'
        
        checkpoint_path = Path(checkpoint_path)
        if not checkpoint_path.exists():
            logger.info("No checkpoint found - starting fresh training")
            return False
        
        logger.info(f"Loading checkpoint: {checkpoint_path}")
        try:
            checkpoint = torch.load(checkpoint_path, map_location=self.device)
            
            # Load model state
            self.model.load_state_dict(checkpoint['model_state_dict'])
            
            # Load optimizer state
            self.optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
            
            # Load scheduler state
            if self.scheduler and checkpoint.get('scheduler_state_dict'):
                self.scheduler.load_state_dict(checkpoint['scheduler_state_dict'])
            
            # Load training state
            self.current_epoch = checkpoint['epoch']
            self.total_steps = checkpoint['total_steps']
            self.best_loss = checkpoint['best_loss']
            self.training_history = checkpoint.get('training_history', [])
            
            logger.info(f"✅ Resumed from epoch {self.current_epoch}, best loss: {self.best_loss:.6f}")
            logger.info(f"Total steps completed: {self.total_steps}")
            return True
            
        except Exception as e:
            logger.error(f"Failed to load checkpoint: {e}")
            return False

    def training_step(self, batch: torch.Tensor) -> Dict[str, float]:
        """Single training step following the ACTUAL model pattern."""
        batch = batch.to(self.device)
        
        # Zero gradients
        self.optimizer.zero_grad()
        
        # Forward pass - EXACTLY like the working basic_training.py
        logits, telemetry = self.model(batch)
        
        # Loss calculation - EXACTLY like example_training_step
        pred = logits[:, :-1, :].reshape(-1, 2)
        target = batch[:, 1:].reshape(-1)
        loss = F.cross_entropy(pred, target)
        
        # Backward pass
        loss.backward()
        
        # Gradient clipping
        torch.nn.utils.clip_grad_norm_(self.model.parameters(), self.config['max_grad_norm'])
        
        # Optimizer step
        self.optimizer.step()
        if self.scheduler:
            self.scheduler.step()
        
        self.total_steps += 1
        
        # Extract telemetry values properly
        metrics = {'loss': loss.item()}
        if telemetry:
            for key, value in telemetry.items():
                if torch.is_tensor(value):
                    metrics[key] = value.mean().item()
                else:
                    metrics[key] = value
        
        return metrics

    def train_epoch(self) -> Dict[str, float]:
        """Train for one complete epoch."""
        logger.info(f"Starting epoch {self.current_epoch + 1}")
        
        # Use EXACT same pattern as working basic_training.py
        self.model.train()
        epoch_losses = []
        
        # Simple batching - EXACTLY like working basic_training.py
        batch_size = self.config['batch_size']
        for i in range(0, len(self.dataset), batch_size):
            batch = self.dataset[i:i + batch_size]
            if len(batch) < batch_size:
                continue  # Skip incomplete batches
            
            batch = batch.to(self.device)
            
            # Zero gradients
            self.optimizer.zero_grad()
            
            # Forward pass - EXACTLY like working basic_training.py
            logits, telemetry = self.model(batch)
            
            # Loss calculation - EXACTLY like working basic_training.py
            pred = logits[:, :-1, :].reshape(-1, 2)
            target = batch[:, 1:].reshape(-1)
            loss = F.cross_entropy(pred, target)
            
            # Backward pass
            loss.backward()
            
            # Gradient clipping
            torch.nn.utils.clip_grad_norm_(self.model.parameters(), self.config['max_grad_norm'])
            
            # Optimizer step
            self.optimizer.step()
            if self.scheduler:
                self.scheduler.step()
            
            self.total_steps += 1
            epoch_losses.append(loss.item())
        
        # Calculate epoch averages - simplified like basic_training.py
        avg_loss = sum(epoch_losses) / len(epoch_losses) if epoch_losses else float('inf')
        
        epoch_summary = {
            'epoch': self.current_epoch + 1,
            'avg_loss': avg_loss
        }
        
        self.training_history.append(epoch_summary)
        
        logger.info(
            f"Epoch {self.current_epoch + 1} completed: "
            f"Avg Loss={avg_loss:.6f}"
        )
        
        return epoch_summary

    def train(self):
        """Main training loop."""
        logger.info("🚀 STARTING BREAKTHROUGH BITRANSFORMERLM TRAINING!")
        logger.info("Configuration: Fixed RL Adafactor + 16M parameters + CPU training")
        
        start_epoch = self.current_epoch
        
        for epoch in range(start_epoch, self.config['num_epochs']):
            try:
                # Train epoch
                epoch_metrics = self.train_epoch()
                avg_loss = epoch_metrics['avg_loss']
                
                # Check if this is the best model
                is_best = avg_loss < self.best_loss
                if is_best:
                    self.best_loss = avg_loss
                
                # Save checkpoint after each epoch
                self.save_checkpoint(self.current_epoch + 1, avg_loss, is_best)
                
                self.current_epoch += 1
                
                # Log progress
                logger.info(f"=== EPOCH {self.current_epoch} COMPLETE ===")
                logger.info(f"Loss: {avg_loss:.6f} (best: {self.best_loss:.6f})")
                
                # Check for breakthrough performance (loss < 3.0)
                if avg_loss < 3.0:
                    logger.info("🚀 BREAKTHROUGH PERFORMANCE ACHIEVED! Loss < 3.0!")
                
            except KeyboardInterrupt:
                logger.info("Training interrupted by user")
                # Save checkpoint before exiting
                try:
                    self.save_checkpoint(self.current_epoch, float('inf'), False)
                except:
                    pass
                break
            except Exception as e:
                logger.error(f"Error in epoch {self.current_epoch + 1}: {e}")
                # Save emergency checkpoint
                try:
                    self.save_checkpoint(self.current_epoch, float('inf'), False)
                except:
                    pass
                raise

def main():
    """Main function to run breakthrough training."""
    
    # BREAKTHROUGH TRAINING CONFIGURATION
    config = {
        # Model parameters (breakthrough configuration)
        'sequence_length': 512,
        
        # Training parameters
        'learning_rate': 1e-3,        # FIXED LR - key to breakthrough!
        'weight_decay': 0.01,
        'batch_size': 4,              # Adjust based on memory
        'num_epochs': 50,             # Full training run
        'max_grad_norm': 1.0,
        
        # Data parameters
        'hf_token': None,  # Set via environment variable HF_TOKEN
        
        # Logging and checkpointing
        'log_interval': 100,
        'checkpoint_dir': '/data/BitTransformerLM/checkpoints',
    }
    
    # Create trainer
    trainer = BreakthroughTrainer(config)
    
    # Setup all components
    logger.info("Setting up training components...")
    trainer.load_and_prepare_dataset()
    trainer.create_breakthrough_model()
    trainer.setup_optimizer()
    
    # Try to resume from checkpoint
    trainer.load_checkpoint()
    
    # Start training
    trainer.train()
    
    logger.info("🎉 BREAKTHROUGH TRAINING COMPLETED!")
    logger.info(f"Best loss achieved: {trainer.best_loss:.6f}")
    logger.info(f"Checkpoints saved to: {trainer.checkpoint_dir}")

if __name__ == "__main__":
    main()