Upload 6 files

e8bf402 verified 12 months ago

7.37 kB

#!/usr/bin/env python3
"""
Example Usage: Shakespeare Transformer
This script shows how to download and use the Shakespeare model from Hugging Face.
Usage:
    python example_usage.py
"""

import torch
import torch.nn as nn

print("="*70)
print("🎭 Shakespeare Transformer - Example Usage")
print("="*70)
print()

# Check device
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f"Device: {device}")
print()

# ============================================
# METHOD 1: Download from Hugging Face
# ============================================

print("📥 Method 1: Download from Hugging Face")
print("-"*70)
print()
print("To download the model:")
print()
print("from huggingface_hub import hf_hub_download")
print()
print("# Download model file")
print("model_path = hf_hub_download(")
print("    repo_id='YOUR-USERNAME/shakespeare-transformer-learning',")
print("    filename='best_model.pth'")
print(")")
print()
print("# Load the model")
print("checkpoint = torch.load(model_path, map_location=device)")
print()

# ============================================
# METHOD 2: Use Local File
# ============================================

print()
print("📂 Method 2: Use Local File")
print("-"*70)
print()

# Define the CharTokenizer class (needed for loading)
class CharTokenizer:
    def __init__(self, text=None):
        if text is not None:
            self.chars = sorted(list(set(text)))
            self.vocab_size = len(self.chars)
            self.char_to_idx = {ch: i for i, ch in enumerate(self.chars)}
            self.idx_to_char = {i: ch for i, ch in enumerate(self.chars)}
        else:
            self.chars = []
            self.vocab_size = 0
            self.char_to_idx = {}
            self.idx_to_char = {}
    
    def encode(self, text):
        return [self.char_to_idx[ch] for ch in text if ch in self.char_to_idx]
    
    def decode(self, indices):
        return ''.join([self.idx_to_char.get(i, '') for i in indices])

# Define the model architecture
class TransformerLanguageModel(nn.Module):
    def __init__(self, vocab_size, d_model=256, nhead=8, num_layers=6, dropout=0.2, seq_length=128):
        super().__init__()
        self.d_model = d_model
        self.seq_length = seq_length
        self.embedding = nn.Embedding(vocab_size, d_model)
        self.pos_encoding = nn.Embedding(seq_length, d_model)
        
        encoder_layer = nn.TransformerEncoderLayer(
            d_model=d_model,
            nhead=nhead,
            dim_feedforward=d_model * 4,
            dropout=dropout,
            batch_first=True
        )
        self.transformer = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
        
        self.dropout = nn.Dropout(dropout)
        self.fc_out = nn.Linear(d_model, vocab_size)
    
    def forward(self, x):
        batch_size, seq_len = x.shape
        token_emb = self.embedding(x)
        positions = torch.arange(seq_len, device=x.device).unsqueeze(0).expand(batch_size, -1)
        pos_emb = self.pos_encoding(positions)
        x = self.dropout(token_emb + pos_emb)
        mask = nn.Transformer.generate_square_subsequent_mask(seq_len).to(x.device)
        x = self.transformer(x, mask=mask, is_causal=True)
        logits = self.fc_out(x)
        return logits

# Load model
try:
    print("Loading model...")
    checkpoint = torch.load('best_model.pth', map_location=device, weights_only=False)
    tokenizer = checkpoint['tokenizer']
    
    model = TransformerLanguageModel(
        vocab_size=tokenizer.vocab_size,
        d_model=256,
        nhead=8,
        num_layers=6,
        dropout=0.2,
        seq_length=128
    ).to(device)
    
    model.load_state_dict(checkpoint['model_state_dict'])
    model.eval()
    
    print("✓ Model loaded successfully!")
    print()
    
except FileNotFoundError:
    print("⚠️  best_model.pth not found in current directory")
    print("Please download it from Hugging Face first.")
    exit()

# ============================================
# GENERATION FUNCTION
# ============================================

def generate_text(prompt, max_length=300, temperature=0.8):
    """
    Generate text from a prompt
    
    Args:
        prompt: Starting text (e.g., "ROMEO:" or "To be or not to be")
        max_length: Maximum number of characters to generate
        temperature: Sampling temperature (higher = more random)
    
    Returns:
        Generated text as string
    """
    model.eval()
    indices = tokenizer.encode(prompt) if prompt else [0]
    
    with torch.no_grad():
        for _ in range(max_length):
            # Get last seq_length characters
            x = torch.tensor(indices[-128:], dtype=torch.long).unsqueeze(0).to(device)
            
            # Pad if needed
            if x.shape[1] < 128:
                padding = torch.zeros(1, 128 - x.shape[1], dtype=torch.long).to(device)
                x = torch.cat([padding, x], dim=1)
            
            # Generate next character
            logits = model(x)
            logits = logits[0, -1, :] / temperature
            probs = torch.softmax(logits, dim=-1)
            next_idx = torch.multinomial(probs, num_samples=1).item()
            indices.append(next_idx)
    
    return tokenizer.decode(indices)

# ============================================
# EXAMPLE GENERATIONS
# ============================================

print("🎬 Example Generations")
print("="*70)
print()

examples = [
    ("ROMEO:", "Character dialogue"),
    ("To be or not to be", "Famous quote continuation"),
    ("Once upon a time", "Story beginning"),
    ("", "Random generation"),
]

for prompt, description in examples:
    print(f"📝 {description}")
    print(f"Prompt: '{prompt}'")
    print("-"*70)
    
    generated = generate_text(prompt, max_length=200, temperature=0.8)
    
    # Show first 300 characters
    display_text = generated[:300]
    if len(generated) > 300:
        display_text += "..."
    
    print(display_text)
    print()
    print("="*70)
    print()

# ============================================
# INTERACTIVE MODE
# ============================================

print("🎮 Interactive Mode")
print("="*70)
print("Enter prompts to generate text. Type 'quit' to exit.")
print()

while True:
    try:
        prompt = input("\nEnter prompt (or 'quit'): ")
        
        if prompt.lower() in ['quit', 'exit', 'q']:
            print("Goodbye! 👋")
            break
        
        print("\nGenerating...")
        print("-"*70)
        
        generated = generate_text(prompt, max_length=300, temperature=0.8)
        print(generated[:400])  # Show first 400 characters
        
        print("-"*70)
        
    except KeyboardInterrupt:
        print("\n\nGoodbye! 👋")
        break

# ============================================
# TIPS
# ============================================

print()
print("💡 Tips for Best Results:")
print("="*70)
print()
print("1. Use character names as prompts: 'ROMEO:', 'JULIET:', etc.")
print("2. Start with famous quotes: 'To be or not to be'")
print("3. Try lower temperature (0.5) for more consistent text")
print("4. Try higher temperature (1.2) for more creative/random text")
print("5. This is a small educational model - expect imperfections!")
print()
print("🎭 Enjoy exploring Shakespeare-style text generation!")

	#!/usr/bin/env python3
	"""
	Example Usage: Shakespeare Transformer

	This script shows how to download and use the Shakespeare model from Hugging Face.

	Usage:
	python example_usage.py
	"""

	import torch
	import torch.nn as nn

	print("="*70)
	print("🎭 Shakespeare Transformer - Example Usage")
	print("="*70)
	print()

	# Check device
	device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
	print(f"Device: {device}")
	print()

	# ============================================
	# METHOD 1: Download from Hugging Face
	# ============================================

	print("📥 Method 1: Download from Hugging Face")
	print("-"*70)
	print()
	print("To download the model:")
	print()
	print("from huggingface_hub import hf_hub_download")
	print()
	print("# Download model file")
	print("model_path = hf_hub_download(")
	print(" repo_id='YOUR-USERNAME/shakespeare-transformer-learning',")
	print(" filename='best_model.pth'")
	print(")")
	print()
	print("# Load the model")
	print("checkpoint = torch.load(model_path, map_location=device)")
	print()

	# ============================================
	# METHOD 2: Use Local File
	# ============================================

	print()
	print("📂 Method 2: Use Local File")
	print("-"*70)
	print()

	# Define the CharTokenizer class (needed for loading)
	class CharTokenizer:
	def __init__(self, text=None):
	if text is not None:
	self.chars = sorted(list(set(text)))
	self.vocab_size = len(self.chars)
	self.char_to_idx = {ch: i for i, ch in enumerate(self.chars)}
	self.idx_to_char = {i: ch for i, ch in enumerate(self.chars)}
	else:
	self.chars = []
	self.vocab_size = 0
	self.char_to_idx = {}
	self.idx_to_char = {}

	def encode(self, text):
	return [self.char_to_idx[ch] for ch in text if ch in self.char_to_idx]

	def decode(self, indices):
	return ''.join([self.idx_to_char.get(i, '') for i in indices])

	# Define the model architecture
	class TransformerLanguageModel(nn.Module):
	def __init__(self, vocab_size, d_model=256, nhead=8, num_layers=6, dropout=0.2, seq_length=128):
	super().__init__()
	self.d_model = d_model
	self.seq_length = seq_length
	self.embedding = nn.Embedding(vocab_size, d_model)
	self.pos_encoding = nn.Embedding(seq_length, d_model)

	encoder_layer = nn.TransformerEncoderLayer(
	d_model=d_model,
	nhead=nhead,
	dim_feedforward=d_model * 4,
	dropout=dropout,
	batch_first=True
	)
	self.transformer = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)

	self.dropout = nn.Dropout(dropout)
	self.fc_out = nn.Linear(d_model, vocab_size)

	def forward(self, x):
	batch_size, seq_len = x.shape
	token_emb = self.embedding(x)
	positions = torch.arange(seq_len, device=x.device).unsqueeze(0).expand(batch_size, -1)
	pos_emb = self.pos_encoding(positions)
	x = self.dropout(token_emb + pos_emb)
	mask = nn.Transformer.generate_square_subsequent_mask(seq_len).to(x.device)
	x = self.transformer(x, mask=mask, is_causal=True)
	logits = self.fc_out(x)
	return logits

	# Load model
	try:
	print("Loading model...")
	checkpoint = torch.load('best_model.pth', map_location=device, weights_only=False)
	tokenizer = checkpoint['tokenizer']

	model = TransformerLanguageModel(
	vocab_size=tokenizer.vocab_size,
	d_model=256,
	nhead=8,
	num_layers=6,
	dropout=0.2,
	seq_length=128
	).to(device)

	model.load_state_dict(checkpoint['model_state_dict'])
	model.eval()

	print("✓ Model loaded successfully!")
	print()

	except FileNotFoundError:
	print("⚠️ best_model.pth not found in current directory")
	print("Please download it from Hugging Face first.")
	exit()

	# ============================================
	# GENERATION FUNCTION
	# ============================================

	def generate_text(prompt, max_length=300, temperature=0.8):
	"""
	Generate text from a prompt

	Args:
	prompt: Starting text (e.g., "ROMEO:" or "To be or not to be")
	max_length: Maximum number of characters to generate
	temperature: Sampling temperature (higher = more random)

	Returns:
	Generated text as string
	"""
	model.eval()
	indices = tokenizer.encode(prompt) if prompt else [0]

	with torch.no_grad():
	for _ in range(max_length):
	# Get last seq_length characters
	x = torch.tensor(indices[-128:], dtype=torch.long).unsqueeze(0).to(device)

	# Pad if needed
	if x.shape[1] < 128:
	padding = torch.zeros(1, 128 - x.shape[1], dtype=torch.long).to(device)
	x = torch.cat([padding, x], dim=1)

	# Generate next character
	logits = model(x)
	logits = logits[0, -1, :] / temperature
	probs = torch.softmax(logits, dim=-1)
	next_idx = torch.multinomial(probs, num_samples=1).item()
	indices.append(next_idx)

	return tokenizer.decode(indices)

	# ============================================
	# EXAMPLE GENERATIONS
	# ============================================

	print("🎬 Example Generations")
	print("="*70)
	print()

	examples = [
	("ROMEO:", "Character dialogue"),
	("To be or not to be", "Famous quote continuation"),
	("Once upon a time", "Story beginning"),
	("", "Random generation"),
	]

	for prompt, description in examples:
	print(f"📝 {description}")
	print(f"Prompt: '{prompt}'")
	print("-"*70)

	generated = generate_text(prompt, max_length=200, temperature=0.8)

	# Show first 300 characters
	display_text = generated[:300]
	if len(generated) > 300:
	display_text += "..."

	print(display_text)
	print()
	print("="*70)
	print()

	# ============================================
	# INTERACTIVE MODE
	# ============================================

	print("🎮 Interactive Mode")
	print("="*70)
	print("Enter prompts to generate text. Type 'quit' to exit.")
	print()

	while True:
	try:
	prompt = input("\nEnter prompt (or 'quit'): ")

	if prompt.lower() in ['quit', 'exit', 'q']:
	print("Goodbye! 👋")
	break

	print("\nGenerating...")
	print("-"*70)

	generated = generate_text(prompt, max_length=300, temperature=0.8)
	print(generated[:400]) # Show first 400 characters

	print("-"*70)

	except KeyboardInterrupt:
	print("\n\nGoodbye! 👋")
	break

	# ============================================
	# TIPS
	# ============================================

	print()
	print("💡 Tips for Best Results:")
	print("="*70)
	print()
	print("1. Use character names as prompts: 'ROMEO:', 'JULIET:', etc.")
	print("2. Start with famous quotes: 'To be or not to be'")
	print("3. Try lower temperature (0.5) for more consistent text")
	print("4. Try higher temperature (1.2) for more creative/random text")
	print("5. This is a small educational model - expect imperfections!")
	print()
	print("🎭 Enjoy exploring Shakespeare-style text generation!")