Update app.py

app.py

@@ -42,7 +42,7 @@ def run_training(hf_token, model_name, new_repo_name, lora_r, lora_alpha, lora_d
     except Exception as e:
         return f"Error de autenticación: {str(e)}"
 
-    device = "cuda" if torch.cuda.is_available() else "cpu"
+#    device = "cuda" if torch.cuda.is_available() else "cpu"
     num_workers = multiprocessing.cpu_count()
 
     if not hasattr(torch, 'xla'):
@@ -166,7 +166,7 @@ def run_training(hf_token, model_name, new_repo_name, lora_r, lora_alpha, lora_d
 
     progress(0.4, desc="Cargando Modelo...")
     try:
-        original_model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True).to(device)
+        original_model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
     except Exception as e:
         return f"Error cargando modelo: {str(e)}"
     
@@ -216,7 +216,7 @@ def run_training(hf_token, model_name, new_repo_name, lora_r, lora_alpha, lora_d
     trainer.save_model(output_dir)
 
     progress(0.9, desc="Fusionando...")
-    ft = PeftModel.from_pretrained(original_model, output_dir, torch_dtype=torch.float32, is_trainable=False, device_map={"": device}).merge_and_unload()
+    ft = PeftModel.from_pretrained(original_model, output_dir, torch_dtype=torch.float32, is_trainable=False).merge_and_unload()
     
     final_path = "/content/merged_model"
     ft.save_pretrained(final_path, safe_serialization=True)