Update app_test.py

app_test.py

@@ -64,6 +64,12 @@ LOGDIR = external_log_dir
 VOTEDIR = "./votes"
 
 
+@spaces.GPU
+def bot(history, temperature, top_p, max_output_tokens):
+    our_chatbot = chat_manager.get_chatbot(args, model_path, tokenizer, model, image_processor, context_len)
+    print(f"### Chatbot instance ID: {id(our_chatbot)}")
+
+
 with gr.Blocks(
     css=".message-wrap.svelte-1lcyrx4>div.svelte-1lcyrx4  img {min-width: 40px}",
 ) as demo:
@@ -104,10 +110,11 @@ with gr.Blocks(
             upvote_btn = gr.Button(value="👍  Upvote", interactive=True)
             downvote_btn = gr.Button(value="👎  Downvote", interactive=True)
             flag_btn = gr.Button(value="⚠️  Flag", interactive=True)
-            # stop_btn = gr.Button(value="⏹️  Stop Generation", interactive=True)
             regenerate_btn = gr.Button(value="🔄  Regenerate", interactive=True)
             clear_btn = gr.Button(value="🗑️  Clear history", interactive=True)
 
+    bot()
+
 demo.queue()
 
 if __name__ == "__main__":
@@ -130,8 +137,8 @@ if __name__ == "__main__":
 
     model_path = args.model_path
     filt_invalid = "cut"
-    model_name = get_model_name_from_path(args.model_path)
-    tokenizer, model, image_processor, context_len = load_pretrained_model(args.model_path, args.model_base, model_name, args.load_8bit, args.load_4bit)
-    model=model.to(torch.device('cuda'))
+    #model_name = get_model_name_from_path(args.model_path)
+    #tokenizer, model, image_processor, context_len = load_pretrained_model(args.model_path, args.model_base, model_name, args.load_8bit, args.load_4bit)
+    #model=model.to(torch.device('cuda'))
     chat_image_num = 0
     demo.launch()