Update app.py

app.py

@@ -42,12 +42,16 @@ def gpt3(question,vqa_answer,caption):
 
     
 def inference_chat(input_image,input_text):
+    cap=caption(input_image)
     inputs = processor(images=input_image, text=input_text,return_tensors="pt")
     inputs["max_length"] = 10
     inputs["num_beams"] = 5
     inputs['num_return_sequences'] =4
     out = model_vqa.generate(**inputs)
-    return "\n".join(processor.batch_decode(out, skip_special_tokens=True))
+    out=processor.batch_decode(out, skip_special_tokens=True)
+    vqa="\n".join(out)
+    gpt3_out=gpt3(input_text,vqa,cap)
+    return out[0], gpt3_out
     
 with gr.Blocks(
     css="""
@@ -70,18 +74,20 @@ with gr.Blocks(
                     with gr.Row():
                         clear_button = gr.Button(value="Clear", interactive=True)
                         submit_button = gr.Button(
-                            value="Submit_VQA", interactive=True, variant="primary"
+                            value="VLE", interactive=True, variant="primary"
                         )
+                        '''
                     cap_submit_button = gr.Button(
                             value="Submit_CAP", interactive=True, variant="primary"
                         )
                     gpt3_submit_button = gr.Button(
                             value="Submit_GPT3", interactive=True, variant="primary"
                         )
+                        '''
         with gr.Column():
-            caption_output = gr.Textbox(lines=0, label="VQA Output(模型答案输出)")
-            caption_output_v1 = gr.Textbox(lines=0, label="Caption Output(模型caption输出)")
-            gpt3_output_v1 = gr.Textbox(lines=0, label="GPT3 Output(GPT3输出)")
+            caption_output = gr.Textbox(lines=0, label="VQA ")
+            #caption_output_v1 = gr.Textbox(lines=0, label="Caption Output(模型caption输出)")
+            gpt3_output_v1 = gr.Textbox(lines=0, label="VQA+LLM")
             
         image_input.change(
             lambda: ("", "", []),
@@ -100,7 +106,7 @@ with gr.Blocks(
         clear_button.click(
                         lambda: ("", [], []),
                         [],
-                        [chat_input,  state],
+                        [chat_input,  state,caption_output,gpt3_output_v1],
                         queue=False,
                     )
         submit_button.click(
@@ -109,8 +115,9 @@ with gr.Blocks(
                             image_input,
                             chat_input,
                         ],
-                        [caption_output],
+                        [caption_output,gpt3_output_v1],
                     )
+        '''
         cap_submit_button.click(
                         caption,
                         [
@@ -128,6 +135,7 @@ with gr.Blocks(
                         ],
                         [gpt3_output_v1],
                     )
+        '''
 
    # examples = gr.Examples(
    #     examples=examples,