Update app.py

app.py

@@ -4,15 +4,18 @@ import torch
 
 def load_model():
     model_name = "TheBloke/Llama-2-13B-chat-GPTQ"
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=True)
     model = AutoModelForCausalLM.from_pretrained(
         model_name,
         device_map="auto",
         trust_remote_code=True,
-        revision="main"
+        revision="main",
+        quantization_config={"load_in_4bit": True}  # Dodajemy kwantyzację 4-bitową
     )
     return model, tokenizer
 
+# Reszta kodu pozostaje bez zmian
+
 def generate_response(prompt, max_length=100):
     try:
         inputs = tokenizer(prompt, return_tensors="pt").to(model.device)