a

app.py

@@ -5,14 +5,14 @@ from huggingface_hub import InferenceClient
 Copied from inference in colab notebook
 """
 
-from transformers import LlamaForCausalLM, LlamaTokenizer
+from transformers import LlamaForCausalLM, AutoTokenizer
 import torch
 
 # Load model and tokenizer globally to avoid reloading for every request
 model_path = "llama_lora_model_1"
 
 # Load tokenizer
-tokenizer = LlamaTokenizer.from_pretrained(model_path)
+tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True, legacy=False)
 
 # Load model
 model = LlamaForCausalLM.from_pretrained(