Create app.py

app.py

@@ -0,0 +1,89 @@
+import os
+import gradio as gr
+from preprocess import process_dataset
+import subprocess
+import time
+
+def train_lora_interface(dataset_zip, model_name, lora_rank, learning_rate, num_epochs, hub_token):
+    # 1. Pré-processamento
+    with gr.Progress() as progress:
+        progress(0, "Descompactando e processando dataset...")
+        dataset_dir = process_dataset(dataset_zip, "processed_data")
+        
+        # 2. Configura treinamento
+        progress(0.3, "Configurando treinamento LoRA...")
+        output_dir = "lora-output"
+        os.makedirs(output_dir, exist_ok=True)
+        
+        # 3. Executa treinamento
+        progress(0.5, "Treinando modelo (isso pode levar horas)...")
+        cmd = [
+            "python", "train_lora.py",
+            "--dataset_dir", dataset_dir,
+            "--model_name", model_name,
+            "--lora_rank", str(lora_rank),
+            "--learning_rate", str(learning_rate),
+            "--num_epochs", str(num_epochs),
+            "--output_dir", output_dir
+        ]
+        
+        if hub_token:
+            os.environ["HF_TOKEN"] = hub_token
+            cmd.append("--push_to_hub")
+            cmd.append("--hub_model_id")
+            cmd.append("my-lora-model")
+        
+        process = subprocess.Popen(
+            cmd, 
+            stdout=subprocess.PIPE, 
+            stderr=subprocess.STDOUT,
+            universal_newlines=True
+        )
+        
+        logs = ""
+        for line in process.stdout:
+            logs += line
+            progress(0.7, f"Treinando...\n{logs[-200:]}")
+        
+        # 4. Finalização
+        progress(0.9, "Subindo para Hugging Face Hub...")
+        if hub_token:
+            from huggingface_hub import upload_folder
+            upload_folder(
+                repo_id="my-lora-model",
+                folder_path=output_dir,
+                token=hub_token
+            )
+        
+        progress(1.0, "Treinamento concluído com sucesso!")
+        return f"Modelo salvo em: {output_dir}\nLogs: {logs[-500:]}"
+
+# Interface Gradio
+with gr.Blocks(title="LoRA Trainer - Hugging Face") as demo:
+    gr.Markdown("# 🚀 Treinador de LoRA para Stable Diffusion")
+    gr.Markdown("Treine seus próprios modelos LoRA diretamente no Hugging Face Spaces!")
+    
+    with gr.Tab("Configuração"):
+        dataset_zip = gr.File(label="Dataset (ZIP com imagens)", file_types=['.zip'])
+        model_name = gr.Dropdown(
+            ["runwayml/stable-diffusion-v1-5", "stabilityai/stable-diffusion-2-1"], 
+            value="runwayml/stable-diffusion-v1-5",
+            label="Modelo Base"
+        )
+        lora_rank = gr.Slider(4, 64, value=4, step=4, label="Rank LoRA")
+        learning_rate = gr.Number(value=1e-4, label="Taxa de Aprendizado")
+        num_epochs = gr.Slider(1, 50, value=10, step=1, label="Épocas")
+        hub_token = gr.Textbox(label="Token Hugging Face (opcional)", type="password")
+    
+    with gr.Tab("Treinamento"):
+        start_btn = gr.Button("🚀 Iniciar Treinamento", variant="primary")
+        output = gr.Textbox(label="Logs do Treinamento")
+    
+    start_btn.click(
+        fn=train_lora_interface,
+        inputs=[dataset_zip, model_name, lora_rank, learning_rate, num_epochs, hub_token],
+        outputs=output
+    )
+
+if __name__ == "__main__":
+    demo.launch()