Simplify demo: use 1 model, reduce default number of images

app.py

@@ -322,26 +322,36 @@ def main():
             header = """
                     # MLR Text-to-Image Diffusion Model Web Demo
 
+                    This is a demo of model `mdm-flickr-64`. For additional models, please check [our repo](https://github.com/apple/ml-mdm).
+
                     ### Usage
-                    - Select examples below or manually input model and prompts
+                    - Select examples below or manually input prompt
                     - Change more advanced settings such as inference steps.
                     """
             gr.Markdown(header)
 
         with gr.Row(equal_height=False):
             pid = gr.State()
-            with gr.Column(scale=2):
+            ckpt_name = gr.Label("mdm-flickr-64", visible=False)
+            # with gr.Row():
+            prompt_input = gr.Textbox(label="Input prompt")
+
+        with gr.Row(equal_height=False):
+            with gr.Column(scale=1):
                 with gr.Row(equal_height=False):
-                    with gr.Column(scale=1):
-                        ckpt_name = gr.Dropdown(
-                            [
-                                "mdm-flickr-64",
-                                "mdm-flickr-256",
-                                "mdm-flickr-1024",
-                            ],
-                            value="mdm-flickr-64",
-                            label="Model",
-                        )
+                    guidance_scale = gr.Slider(
+                        value=7.5,
+                        minimum=0.0,
+                        maximum=50,
+                        step=0.1,
+                        label="Guidance scale",
+                    )
+                with gr.Row(equal_height=False):
+                    batch_size = gr.Slider(
+                        value=4, minimum=1, maximum=32, step=1, label="Number of images"
+                    )
+
+            with gr.Column(scale=1):
                 with gr.Row(equal_height=False):
                     with gr.Column(scale=1):
                         save_diffusion_path = gr.Checkbox(
@@ -350,27 +360,12 @@ def main():
                         show_diffusion_path = gr.Checkbox(
                             value=False, label="Show diffusion progress"
                         )
-                    with gr.Column(scale=1):
                         show_xt = gr.Checkbox(value=False, label="Show predicted x_t")
+                    with gr.Column(scale=1):
                         output_inner = gr.Checkbox(
                             value=False,
                             label="Output inner UNet (High-res models Only)",
-                        )
-
-            with gr.Column(scale=2):
-                prompt_input = gr.Textbox(label="Input prompt")
-                with gr.Row(equal_height=False):
-                    with gr.Column(scale=1):
-                        guidance_scale = gr.Slider(
-                            value=7.5,
-                            minimum=0.0,
-                            maximum=50,
-                            step=0.1,
-                            label="Guidance scale",
-                        )
-                    with gr.Column(scale=1):
-                        batch_size = gr.Slider(
-                            value=64, minimum=1, maximum=128, step=1, label="Number of images"
+                            visible=False,
                         )
 
         with gr.Row(equal_height=False):
@@ -530,14 +525,14 @@ def main():
             cancels=[run_event],
             queue=False,
         )
-        example0 = gr.Examples(
-            [
-                ["mdm-flickr-64", 64, 50, 0],
-                ["mdm-flickr-256", 16, 100, 0],
-                ["mdm-flickr-1024", 4, 250, 1],
-            ],
-            inputs=[ckpt_name, batch_size, num_inference_steps, eta],
-        )
+        # example0 = gr.Examples(
+        #     [
+        #         ["mdm-flickr-64", 64, 50, 0],
+        #         ["mdm-flickr-256", 16, 100, 0],
+        #         ["mdm-flickr-1024", 4, 250, 1],
+        #     ],
+        #     inputs=[ckpt_name, batch_size, num_inference_steps, eta],
+        # )
         example1 = gr.Examples(
             examples=[[t.strip()] for t in example_texts],
             inputs=[prompt_input],