Update app.py

app.py

@@ -4,7 +4,11 @@ from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 # Carregar o modelo direto do HuggingFace Hub
 model_id = "microsoft/phi-2"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",
+    offload_folder="offload"
+)
 
 # Pipeline de geração de texto
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)