Update app.py

app.py

@@ -1,3 +1,4 @@
+import os, textwrap, torch, gradio as gr
 import torch, textwrap, gradio as gr
 from transformers import (
     AutoTokenizer,
@@ -5,15 +6,17 @@ from transformers import (
     BitsAndBytesConfig,
     pipeline,
 )
+AUTH = os.environ.get("HF_TOKEN")  
 MODEL_ID = "mistralai/Mixtral-8x7B-Instruct-v0.1"   # FP16 weights
 bnb_cfg  = BitsAndBytesConfig(
     load_in_4bit=True,
     bnb_4bit_compute_dtype=torch.float16,   # keeps mat-mul fast
 )
 
-tok   = AutoTokenizer.from_pretrained(MODEL_ID, use_fast=True)
+tok   = AutoTokenizer.from_pretrained(MODEL_ID, token=AUTH, use_fast=True)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
+    token=AUTH,
     device_map="auto",
     trust_remote_code=True,
     quantization_config=bnb_cfg,            # perfectly fine here