Add clear last round button.

.gitignore

@@ -1,2 +1,4 @@
 .DS_Store
 venv/
+__pycache__
+*.pyc

app.py

@@ -52,6 +52,13 @@ def reset():
     return [[], [], None], []
 
 
+def reset_last(state):
+    conversation = state[0][:-1]
+    chat_history = state[1][:-2]
+    input_image = state[2]
+    return [conversation, chat_history, input_image], conversation
+
+
 def save_image_to_local(image: Image.Image):
     # TODO(jykoh): Update so the url path is used, to prevent repeat saving.
     filename = next(tempfile._get_candidate_names()) + '.png'
@@ -81,7 +88,7 @@ def generate_for_prompt(input_text, state, ret_scale_factor, max_nm_rets, num_wo
 
     print('Running model.generate_for_images_and_texts with', model_inputs, flush=True)
     model_outputs = model.generate_for_images_and_texts(model_inputs, 
-        num_words=num_words, ret_scale_factor=ret_scale_factor, top_p=top_p,
+        num_words=max(num_words, 1), ret_scale_factor=ret_scale_factor, top_p=top_p,
         temperature=temperature, max_num_rets=max_nm_rets)
     print('model_outputs', model_outputs, flush=True)
 
@@ -130,17 +137,19 @@ with gr.Blocks(css=css) as demo:
         with gr.Column(scale=0.3, min_width=100):
             ret_scale_factor = gr.Slider(minimum=0.0, maximum=3.0, value=1.0, step=0.1, interactive=True, label="Multiplier for returning images (higher means more frequent)")
             max_ret_images = gr.Number(minimum=0, maximum=3, value=1, precision=1, interactive=True, label="Max images to return")
-            gr_max_len = gr.Number(value=32, precision=1, label="Max # of words returned", interactive=True)
-            gr_temperature = gr.Number(value=0.0, label="Temperature", interactive=True)
+            gr_max_len = gr.Slider(minimum=1, maximum=64, value=32, step=1, interactive=True, label="Max # of words returned")
+            gr_temperature = gr.Slider(minimum=0.0, maximum=1.0, value=0.0, interactive=True, label="Temperature")
 
         with gr.Column(scale=0.7, min_width=400):
             image_btn = gr.UploadButton("🖼️ Image Input", file_types=["image"])
             text_input = gr.Textbox(label="Chat Input", lines=1, placeholder="Upload an image above [optional]. Then enter a text prompt, and press enter!")
 
             with gr.Row():
-                with gr.Column(scale=0.5):
+                with gr.Column(scale=0.33):
                     submit_btn = gr.Button("Submit", interactive=True, variant="primary")
-                with gr.Column(scale=0.5):
+                with gr.Column(scale=0.33):
+                    clear_last_btn = gr.Button("Clear Last Round")
+                with gr.Column(scale=0.33):
                     clear_btn = gr.Button("Clear History")
 
     text_input.submit(generate_for_prompt, [text_input, gr_state, ret_scale_factor, max_ret_images, gr_max_len, gr_temperature], [gr_state, chatbot])
@@ -149,9 +158,10 @@ with gr.Blocks(css=css) as demo:
     submit_btn.click(lambda: "", None, text_input)  # Reset chatbox.
 
     image_btn.upload(upload_image, [gr_state, image_btn], [gr_state, chatbot])
+    clear_last_btn.click(reset_last, [gr_state], [gr_state, chatbot])
     clear_btn.click(reset, [], [gr_state, chatbot])
     share_button.click(None, [], [], _js=share_js)
 
 
-demo.queue(concurrency_count=1, api_open=False, max_size=16)
-demo.launch(debug=True, server_name="0.0.0.0")
+# demo.queue(concurrency_count=1, api_open=False, max_size=16)
+demo.launch(debug=True, server_name="127.0.0.1")

fromage/models.py

@@ -635,7 +635,7 @@ def load_fromage(embeddings_dir: str, model_args_path: str, model_ckpt_path: str
   assert len(ret_token_idx) == 1, ret_token_idx
   model_kwargs['retrieval_token_idx'] = ret_token_idx[0]
 
-  debug = False
+  debug = True
   if debug:
     model_kwargs['opt_version'] = 'facebook/opt-125m'
     model_kwargs['visual_encoder'] = 'openai/clip-vit-base-patch32'