debug

app.py

@@ -182,7 +182,27 @@ def http_bot(state):
         state.messages[-1][-1] = server_error_msg
         yield (state, state.to_gradio_chatbot_public()) + (enable_btn,) * 2
         return
+    state.messages[-1][-1] = state.messages[-1][-1][:-1]
+    yield (state, state.to_gradio_chatbot_public()) + (disable_btn,) * 2
     
+    # Step 2: Query-conditioned Caption
+    state.append_message(state.roles[1], "# Query-conditioned Caption\n\n▌")
+    try:
+        for generated_text in stream_response(mllm, cap_inputs, mllm_streamer, cao_prompt, mllm_sampling):
+            output = generated_text[len(cap_prompt):].strip()
+            state.messages[-1][-1] = "# Query-conditioned Caption\n\n" + output + "▌"
+            yield (state, state.to_gradio_chatbot_public()) + (disable_btn,) * 2
+    except Exception as e:
+        os.system("nvidia-smi")
+        logging.info(traceback.print_exc())
+        state.messages[-1][-1] = server_error_msg
+        yield (state, state.to_gradio_chatbot_public()) + (enable_btn,) * 2
+        return
+    state.messages[-1][-1] = state.messages[-1][-1][:-1]
+    yield (state, state.to_gradio_chatbot_public()) + (disable_btn,) * 2
+
+
+
     # caption_text = run_mllm_caption(image_tensor, cap_prompt, qa_prompt)
     # state.append_message(state.roles[1], "# Caption\n\n" + caption_text)
     # logging.info("# Caption\n\n" + caption_text)