Update app.py

app.py

@@ -1,24 +1,20 @@
-import os, textwrap, torch, gradio as gr
-from transformers import (
-    AutoTokenizer,
-    AutoModelForCausalLM,
-    BitsAndBytesConfig,
-    pipeline,
-)
-
-MODEL_ID = "TheBloke/Mixtral-8x7B-Instruct-v0.1-GPTQ"   # ← 8 GB quantised
-bnb_cfg  = BitsAndBytesConfig(load_in_4bit=True)
+from transformers import BitsAndBytesConfig
 
-tok = AutoTokenizer.from_pretrained(MODEL_ID, use_fast=True)
+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)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
     device_map="auto",
     trust_remote_code=True,
-    quantization_config=bnb_cfg,
+    quantization_config=bnb_cfg,            # perfectly fine here
 )
 
-prompt_tmpl = (
+prompt_tpl = (
     "Summarise the following transcript in short in 1 or 2 paragraph and point wise and don't miss any key information cover all"
 )