Update app.py

app.py

@@ -48,12 +48,7 @@ quant_config = BitsAndBytesConfig(
 HF_TOKEN = os.getenv("HF_TOKEN")
 
 def load_models():
-    ner_tok = AutoTokenizer.from_pretrained(
-    NER_ID,
-    token=HF_TOKEN,
-    use_fast=False         
-)
-
+    ner_tok = AutoTokenizer.from_pretrained(NER_ID, token=HF_TOKEN)
     ner_mod = AutoModelForTokenClassification.from_pretrained(NER_ID, token=HF_TOKEN)
     ner_mod.eval()
     if torch.cuda.is_available():
@@ -61,9 +56,8 @@ def load_models():
 
     base_mod = AutoModelForCausalLM.from_pretrained(
         BASE_ID,
-        token=HF_TOKEN,
-        device_map="auto",           
-        offload_folder="/tmp/offload"
+        device_map="auto",
+        token=HF_TOKEN
     )
     norm_tok = AutoTokenizer.from_pretrained(ADAPTER_ID, use_fast=True, token=HF_TOKEN)
     norm_mod = PeftModel.from_pretrained(
@@ -234,5 +228,4 @@ with gr.Blocks() as demo:
     )
 
     # Lanzar la app
-    demo.launch()
-
+    demo.launch()