Update app.py

app.py

@@ -53,8 +53,7 @@ Assistant: girlfriend;mother;father;friend
 torch.set_grad_enabled(False)
 device = "cpu"
 model_name = "TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ"
-token = os.environ['hf_token']
-pipe = pipeline("text-generation", model=model_name, device=device)
+pipe = pipeline("text-generation", model=model_name, device='cuda')
 generate_kwargs = {'max_new_tokens': 20}
 
 # '''
@@ -123,6 +122,7 @@ def generate(text, past_key_values):
 if __name__ == "__main__":
     with torch.no_grad():
         past_key_values = set_past_key_values()
+        pipe.model = pipe.model.cpu()
         demo = gr.Interface(partial(generate, past_key_values=past_key_values), 
                             inputs="textbox", outputs="textbox")
         demo.launch()