Update app.py

app.py

@@ -2,6 +2,7 @@ import os
 from collections.abc import Iterator
 from threading import Thread
 
+import compressed_tensors
 import gradio as gr
 import spaces
 import torch
@@ -25,7 +26,7 @@ if not torch.cuda.is_available():
 
 if torch.cuda.is_available():
     model_id = "shisa-ai/shisa-v2-unphi-14b-W8A8-INT8"
-    model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto", load_in_8bit=True)
+    model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto", load_in_8bit=True, trust_remote_code=True)
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.use_default_system_prompt = False