Update app.py

app.py

@@ -2,6 +2,13 @@ from transformers import TextIteratorStreamer
 from threading import Thread
 import gradio as gr
 
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+checkpoint_id = "sanchit-gandhi/Mistral-7B-Instruct-v0.1"
+
+model = AutoModelForCausalLM.from_pretrained(checkpoint_id, low_cpu_mem_usage=True, load_in_4bit=True)
+tokenizer = AutoTokenizer.from_pretrained(checkpoint_id)
+
 
 MAX_INPUT_TOKEN_LENGTH = 4096