Create app.py

app.py

@@ -0,0 +1,83 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import gradio as gr
+import os
+
+# --- КОНФИГУРАЦИЯ ---
+BLOCK_SIZE = 64
+EMBED_SIZE = 64
+HEADS = 4
+MODEL_PATH = 'minigpt_checkpoint.pt'
+
+# --- АРХИТЕКТУРА ---
+class MiniGPT(nn.Module):
+    def __init__(self, vocab_size, embed_size, num_heads, block_size):
+        super().__init__()
+        self.block_size = block_size
+        self.embedding = nn.Embedding(vocab_size, embed_size)
+        self.pos_embedding = nn.Embedding(block_size, embed_size)
+        encoder_layer = nn.TransformerEncoderLayer(d_model=embed_size, nhead=num_heads, batch_first=True)
+        self.transformer = nn.TransformerEncoder(encoder_layer, num_layers=2)
+        self.fc_out = nn.Linear(embed_size, vocab_size)
+
+    def forward(self, x):
+        B, T = x.shape
+        pos = torch.arange(T, device=x.device).unsqueeze(0)
+        out = self.embedding(x) + self.pos_embedding(pos)
+        out = self.transformer(out)
+        return self.fc_out(out)
+
+# --- ДАННЫЕ И ТОКЕНИЗАЦИЯ ---
+# (В продакшене лучше сохранять словарь в JSON, здесь - упрощенно)
+FILE_NAME = 'book.txt'
+if os.path.exists(FILE_NAME):
+    with open(FILE_NAME, 'r', encoding='utf-8') as f: text = f.read()
+else:
+    text = "привет как дела нормально пока" * 100
+
+chars = sorted(list(set(text)))
+vocab_size = len(chars)
+stoi = { ch:i for i,ch in enumerate(chars) }
+itos = { i:ch for i,ch in enumerate(chars) }
+encode = lambda s: [stoi.get(c, 0) for c in s] # 0 как fallback
+decode = lambda l: ''.join([itos[i] for i in l])
+
+# --- ЗАГРУЗКА МОДЕЛИ ---
+model = MiniGPT(vocab_size, EMBED_SIZE, HEADS, BLOCK_SIZE)
+if os.path.exists(MODEL_PATH):
+    model.load_state_dict(torch.load(MODEL_PATH, map_location=torch.device('cpu')))
+model.eval()
+
+# --- ЛОГИКА ГЕНЕРАЦИИ ---
+def predict(prompt, max_length=50):
+    if not prompt: return "Введите текст"
+    
+    # Ограничиваем входной контекст
+    context_tokens = encode(prompt)[-BLOCK_SIZE:]
+    context = torch.tensor(context_tokens, dtype=torch.long).unsqueeze(0)
+    
+    generated = []
+    for _ in range(max_length):
+        cond = context[:, -BLOCK_SIZE:]
+        with torch.no_grad():
+            logits = model(cond)[:, -1, :]
+            probs = F.softmax(logits, dim=-1)
+            next_token = torch.multinomial(probs, num_samples=1) # Для разнообразия
+            context = torch.cat((context, next_token), dim=1)
+            generated.append(next_token.item())
+            
+    return decode(generated)
+
+# --- ИНТЕРФЕЙС GRADIO ---
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🤖 MiniGPT Chat")
+    with gr.Row():
+        input_text = gr.Textbox(label="Ваш запрос", placeholder="Напишите начало фразы...")
+        output_text = gr.Textbox(label="Ответ модели")
+    
+    btn = gr.Button("Сгенерировать")
+    btn.click(fn=predict, inputs=[input_text], outputs=[output_text])
+
+if __name__ == "__main__":
+    demo.launch()