app.py

import torch
import torch.nn as nn
from torch.nn import functional as F
import tiktoken
import gradio as gr
import torch
import torch.nn as nn
from torch.nn import functional as F
import tiktoken
import gradio as gr
import asyncio 

# Add the post-processing function here
def post_process_text(text):
    # Ensure the text starts with a capital letter
    text = text.capitalize()
    
    # Remove any incomplete sentences at the end
    sentences = text.split('.')
    complete_sentences = sentences[:-1] if len(sentences) > 1 else sentences
    
    # Rejoin sentences and add a period if missing
    processed_text = '. '.join(complete_sentences)
    if not processed_text.endswith('.'):
        processed_text += '.'
    
    return processed_text
# Define the model architecture
class GPTConfig:
    def __init__(self):
        self.block_size = 1024
        self.vocab_size = 50304
        self.n_layer = 12
        self.n_head = 12
        self.n_embd = 768

class CausalSelfAttention(nn.Module):
    def __init__(self, config):
        super().__init__()
        assert config.n_embd % config.n_head == 0
        self.c_attn = nn.Linear(config.n_embd, 3 * config.n_embd)
        self.c_proj = nn.Linear(config.n_embd, config.n_embd)
        self.n_head = config.n_head
        self.n_embd = config.n_embd
        self.register_buffer("bias", torch.tril(torch.ones(config.block_size, config.block_size)).view(1, 1, config.block_size, config.block_size))

    def forward(self, x):
        B, T, C = x.size()
        q, k, v = self.c_attn(x).split(self.n_embd, dim=2)
        k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        q = q.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        v = v.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        y = F.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=0.0, is_causal=True)
        y = y.transpose(1, 2).contiguous().view(B, T, C)
        return self.c_proj(y)

class MLP(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.c_fc = nn.Linear(config.n_embd, 4 * config.n_embd)
        self.gelu = nn.GELU()
        self.c_proj = nn.Linear(4 * config.n_embd, config.n_embd)

    def forward(self, x):
        return self.c_proj(self.gelu(self.c_fc(x)))

class Block(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.ln_1 = nn.LayerNorm(config.n_embd)
        self.attn = CausalSelfAttention(config)
        self.ln_2 = nn.LayerNorm(config.n_embd)
        self.mlp = MLP(config)

    def forward(self, x):
        x = x + self.attn(self.ln_1(x))
        x = x + self.mlp(self.ln_2(x))
        return x

class GPT(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.transformer = nn.ModuleDict(dict(
            wte = nn.Embedding(config.vocab_size, config.n_embd),
            wpe = nn.Embedding(config.block_size, config.n_embd),
            h = nn.ModuleList([Block(config) for _ in range(config.n_layer)]),
            ln_f = nn.LayerNorm(config.n_embd),
        ))
        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
        self.transformer.wte.weight = self.lm_head.weight
        self.apply(self._init_weights)

    def _init_weights(self, module):
        if isinstance(module, nn.Linear):
            torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)
            if module.bias is not None:
                torch.nn.init.zeros_(module.bias)
        elif isinstance(module, nn.Embedding):
            torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)

    def forward(self, idx, targets=None):
        device = idx.device
        b, t = idx.size()
        assert t <= self.config.block_size, f"Cannot forward sequence of length {t}, block size is only {self.config.block_size}"
        pos = torch.arange(0, t, dtype=torch.long, device=device).unsqueeze(0)
        
        tok_emb = self.transformer.wte(idx)
        pos_emb = self.transformer.wpe(pos)
        x = tok_emb + pos_emb
        for block in self.transformer.h:
            x = block(x)
        x = self.transformer.ln_f(x)
        logits = self.lm_head(x)
        
        loss = None
        if targets is not None:
            loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1), ignore_index=-1)
        
        return logits, loss

# Load the model
def load_model(model_path):
    config = GPTConfig()
    model = GPT(config)
    
    checkpoint = torch.load(model_path, map_location=torch.device('cpu'))
    
    print("Checkpoint keys:", checkpoint.keys())  # Debug print
    
    if 'model_state_dict' in checkpoint:
        model.load_state_dict(checkpoint['model_state_dict'])
    else:
        model.load_state_dict(checkpoint)
    
    model.eval()
    return model

# Load the model
model = load_model('gpt_model.pth')  # Replace with the actual path to your .pt file
enc = tiktoken.get_encoding('gpt2')

# Improved text generation function
import torch
import torch.nn as nn
from torch.nn import functional as F
import tiktoken
import gradio as gr

# [Your existing model code remains unchanged]

# Modify the generate_text function to be asynchronous
async def generate_text(prompt, max_length=432, temperature=0.8, top_k=40):
    input_ids = torch.tensor(enc.encode(prompt)).unsqueeze(0)
    generated = []
    
    with torch.no_grad():
        for _ in range(max_length):
            outputs, _ = model(input_ids)
            next_token_logits = outputs[:, -1, :]
            next_token_logits = next_token_logits / temperature
            top_k_logits, top_k_indices = torch.topk(next_token_logits, top_k, dim=-1)
            next_token_probs = F.softmax(top_k_logits, dim=-1)
            next_token_index = torch.multinomial(next_token_probs, num_samples=1)
            next_token = top_k_indices.gather(-1, next_token_index)
            
            input_ids = torch.cat([input_ids, next_token], dim=-1)
            generated.append(next_token.item())
            
            next_token_str = enc.decode([next_token.item()])
            yield next_token_str
            
            if next_token.item() == enc.encode('\n')[0] and len(generated) > 100:
                break
            
            await asyncio.sleep(0.02)  # Slightly faster typing effect

    if len(generated) == max_length:
        yield "... (output truncated due to length)"
# Modify the gradio_generate function to be asynchronous
async def gradio_generate(prompt, max_length, temperature, top_k):
    output = ""
    async for token in generate_text(prompt, max_length, temperature, top_k):
        output += token
        yield output

# Custom CSS for the animation effect
css = """
<style>
    body { background-color: #1e1e1e; color: #ffffff; font-family: Arial, sans-serif; }
    .container { max-width: 800px; margin: 0 auto; padding: 20px; }
    .header { text-align: center; margin-bottom: 30px; }
    .chat-box { background-color: #2a2a2a; border-radius: 10px; padding: 20px; margin-bottom: 20px; }
    .user-input { background-color: #3a3a3a; border: none; color: #ffffff; padding: 10px; border-radius: 5px; width: 100%; }
    .generate-btn { background-color: #5465ff; color: white; border: none; padding: 10px 20px; border-radius: 5px; cursor: pointer; }
    .output-box { background-color: #2a2a2a; border-radius: 10px; padding: 20px; margin-top: 20px; min-height: 100px; }
</style>
"""

with gr.Blocks(css=css) as demo:
    gr.HTML("<div class='header'><h1>🌟 GPT-2 Storyteller</h1></div>")
    
    with gr.Row():
        with gr.Column(scale=3):
            prompt = gr.Textbox(
                placeholder="Start your story here (e.g., 'Once upon a time in a magical forest...')",
                label="Story Prompt",
                elem_classes="user-input"
            )
        with gr.Column(scale=1):
            generate_btn = gr.Button("Generate Story", elem_classes="generate-btn")
    
    with gr.Row():
        max_length = gr.Slider(minimum=50, maximum=500, value=432, step=1, label="Max Length")
        temperature = gr.Slider(minimum=0.1, maximum=1.0, value=0.8, step=0.1, label="Temperature")
        top_k = gr.Slider(minimum=1, maximum=100, value=40, step=1, label="Top-k")
    
    output = gr.Markdown(elem_classes="output-box")
    
    generate_btn.click(
        gradio_generate,
        inputs=[prompt, max_length, temperature, top_k],
        outputs=output
    )


if __name__ == "__main__":
    demo.launch()