Update app.py

app.py

@@ -18,7 +18,8 @@ def load_model():
             bnb_4bit_use_double_quant=True
         )
         
-            model_name = "HuggingFaceH4/zephyr-7b-beta"  # Alternativa compatible
+        model = AutoModelForCausalLM.from_pretrained(
+            "HuggingFaceH4/zephyr-7b-beta",
             device_map="mps",  # Forzar uso de Metal
             quantization_config=quantization_config,
             torch_dtype=torch.bfloat16,  # Tipo de dato compatible