Fix bug with input_image

app.py

@@ -64,19 +64,18 @@ def upload_image(state, image_input):
     input_image = Image.open(image_input.name).resize(
         (224, 224)).convert('RGB')
     input_image.save(image_input.name)  # Overwrite with smaller image.
-    conversation += [(f"![](/file={image_input.name})", "")]
-    return [conversation, chat_history, input_image], conversation
+    conversation += [(f'<img src="/file={image_input.name}" style="display: inline-block;">', "")]
+    return [conversation, chat_history + [input_image, ""]], conversation
 
 
 def reset():
-    return [[], [], None], []
+    return [[], []], []
 
 
 def reset_last(state):
     conversation = state[0][:-1]
     chat_history = state[1][:-2]
-    input_image = state[2]
-    return [conversation, chat_history, input_image], conversation
+    return [conversation, chat_history], conversation
 
 
 def save_image_to_local(image: Image.Image):
@@ -94,16 +93,10 @@ def generate_for_prompt(input_text, state, ret_scale_factor, max_num_rets, num_w
     input_prompt = 'Q: ' + input_text + '\nA:'
     conversation = state[0]
     chat_history = state[1]
-    input_image = state[2]
     print('Generating for', chat_history, flush=True)
 
     # If an image was uploaded, prepend it to the model.
-    model_inputs = None
-    if input_image is not None:
-        model_inputs = chat_history + [input_image]
-    else:
-        model_inputs = chat_history
-
+    model_inputs = chat_history
     model_inputs.append(input_prompt)
 
     top_p = 1.0
@@ -144,8 +137,8 @@ def generate_for_prompt(input_text, state, ret_scale_factor, max_num_rets, num_w
 
     # Set input image to None.
     print('state', state, flush=True)
-    print('updated state', [conversation, chat_history, None], flush=True)
-    return [conversation, chat_history, None], conversation, gr.update(visible=True), gr.update(visible=True)
+    print('updated state', [conversation, chat_history], flush=True)
+    return [conversation, chat_history], conversation, gr.update(visible=True), gr.update(visible=True)
 
 
 with gr.Blocks(css=css) as demo:
@@ -157,7 +150,7 @@ with gr.Blocks(css=css) as demo:
         For faster inference without waiting in queue, you may duplicate the space and use your own GPU. <a href="https://huggingface.co/spaces/haoheliu/audioldm-text-to-audio-generation?duplicate=true"><img style="margin-top: 0em; margin-bottom: 0em" src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a>
     """)
 
-    gr_state = gr.State([[], [], None])  # chat_history, input_image
+    gr_state = gr.State([[], []])  # conversation, chat_history
 
     with gr.Row():
         with gr.Column(scale=0.7, min_width=500):