Update app.py

app.py

@@ -48,12 +48,17 @@ class ModelManager:
         
     def get_llama_pipeline(self, model_id, token):
         if model_id not in self.llama_pipelines:
-            tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=token)
+            tokenizer = AutoTokenizer.from_pretrained(
+                model_id,
+                use_auth_token=token,
+                legacy=False  # Important for compatibility
+            )
             model = AutoModelForCausalLM.from_pretrained(
                 model_id,
                 use_auth_token=token,
                 torch_dtype=torch.float16,
-                device_map="auto"
+                device_map="auto",
+                low_cpu_mem_usage=True  # Reduces memory pressure
             )
             self.llama_pipelines[model_id] = pipeline(
                 "text-generation",