add

app.py

@@ -287,175 +287,178 @@ css = """
     max-width: 640px;
 }
 """
-with gr.Blocks(css=css) as demo:
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown("# CrossFlow")
-        gr.Markdown("[CrossFlow](https://cross-flow.github.io/) directly transforms text representations into images for text-to-image generation, without the need for both the noise distribution and conditioning mechanism.")
-        gr.Markdown("This direct mapping enables meaningful 'Linear Interpolation' and 'Arithmetic Operations' in the text latent space, as demonstrated here.")
-    #     with gr.Tabs():
-    #         with gr.Tab("Linear Interpolation"):
-    #             gr.Markdown("This demo uses 256px images, 25 sampling steps (instead of 50), and 10 interpolations (instead of 50) to conserve GPU memory. For better results, see the original [code](https://github.com/qihao067/CrossFlow). (You may adjust them in Advanced Settings, but doing so may trigger OOM errors.)")
-    #             # gr.Markdown("CrossFlow directly transforms text representations into images for text-to-image generation, enabling interpolation in the input text latent space.")
-
-    #             with gr.Row():
-    #                 prompt1 = gr.Text(
-    #                     label="Prompt_1",
-    #                     show_label=False,
-    #                     max_lines=1,
-    #                     placeholder="Enter your prompt for the first image",
-    #                     container=False,
-    #                 )
+# with gr.Blocks(css=css) as demo:
+#     with gr.Column(elem_id="col-container"):
+#         gr.Markdown("# CrossFlow")
+#         gr.Markdown("[CrossFlow](https://cross-flow.github.io/) directly transforms text representations into images for text-to-image generation, without the need for both the noise distribution and conditioning mechanism.")
+#         gr.Markdown("This direct mapping enables meaningful 'Linear Interpolation' and 'Arithmetic Operations' in the text latent space, as demonstrated here.")
+#         with gr.Tabs():
+#             with gr.Tab("Linear Interpolation"):
+#                 gr.Markdown("This demo uses 256px images, 25 sampling steps (instead of 50), and 10 interpolations (instead of 50) to conserve GPU memory. For better results, see the original [code](https://github.com/qihao067/CrossFlow). (You may adjust them in Advanced Settings, but doing so may trigger OOM errors.)")
+#                 # gr.Markdown("CrossFlow directly transforms text representations into images for text-to-image generation, enabling interpolation in the input text latent space.")
+
+#                 with gr.Row():
+#                     prompt1 = gr.Text(
+#                         label="Prompt_1",
+#                         show_label=False,
+#                         max_lines=1,
+#                         placeholder="Enter your prompt for the first image",
+#                         container=False,
+#                     )
                 
-    #             with gr.Row():
-    #                 prompt2 = gr.Text(
-    #                     label="Prompt_2",
-    #                     show_label=False,
-    #                     max_lines=1,
-    #                     placeholder="Enter your prompt for the second image",
-    #                     container=False,
-    #                 )
-
-    #             with gr.Row():
-    #                 run_button = gr.Button("Run", scale=0, variant="primary")
-
-    #             # Create separate outputs for the first image, last image, and the animated GIF
-    #             first_image_output = gr.Image(label="Image of the first prompt", show_label=True)
-    #             last_image_output = gr.Image(label="Image of the second prompt", show_label=True)
-    #             gif_output = gr.Image(label="Linear interpolation", show_label=True)
-
-    #             with gr.Accordion("Advanced Settings", open=False):
-    #                 seed = gr.Slider(
-    #                     label="Seed",
-    #                     minimum=0,
-    #                     maximum=MAX_SEED,
-    #                     step=1,
-    #                     value=0,
-    #                 )
-
-    #                 randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-
-    #                 with gr.Row():
-    #                     guidance_scale = gr.Slider(
-    #                         label="Guidance scale",
-    #                         minimum=0.0,
-    #                         maximum=10.0,
-    #                         step=0.1,
-    #                         value=7.0,  # Replace with defaults that work for your model
-    #                     )
-    #                 with gr.Row():
-    #                     num_inference_steps = gr.Slider(
-    #                         label="Number of inference steps - 50 inference steps are recommended; but you can reduce to 20 if the demo fails.",
-    #                         minimum=1,
-    #                         maximum=50,
-    #                         step=1,
-    #                         value=25,  # Replace with defaults that work for your model
-    #                     )
-    #                 with gr.Row():
-    #                     num_of_interpolation = gr.Slider(
-    #                         label="Number of images for interpolation - More images yield smoother transitions but require more resources and may fail.",
-    #                         minimum=5,
-    #                         maximum=50,
-    #                         step=1,
-    #                         value=10,  # Replace with defaults that work for your model
-    #                     )
-
-    #             gr.Examples(examples=examples, inputs=[prompt1, prompt2])
-
-    #         with gr.Tab("Arithmetic Operations"):
-    #             # The second tab is currently empty. You can add more components later.
-    #             gr.Markdown("This demo only supports addition or subtraction between two text latents ('Prompt_1 + Prompt_2' or 'Prompt_1 - Prompt_2'). For the other arithmetic operations, see the original [code](https://github.com/qihao067/CrossFlow).")
-    #             with gr.Row():
-    #                 prompt1 = gr.Text(
-    #                     label="Prompt_1",
-    #                     show_label=False,
-    #                     max_lines=1,
-    #                     placeholder="Enter your prompt for the first image",
-    #                     container=False,
-    #                 )
+#                 with gr.Row():
+#                     prompt2 = gr.Text(
+#                         label="Prompt_2",
+#                         show_label=False,
+#                         max_lines=1,
+#                         placeholder="Enter your prompt for the second image",
+#                         container=False,
+#                     )
+
+#                 with gr.Row():
+#                     run_button = gr.Button("Run", scale=0, variant="primary")
+
+#                 # Create separate outputs for the first image, last image, and the animated GIF
+#                 first_image_output = gr.Image(label="Image of the first prompt", show_label=True)
+#                 last_image_output = gr.Image(label="Image of the second prompt", show_label=True)
+#                 gif_output = gr.Image(label="Linear interpolation", show_label=True)
+
+#                 with gr.Accordion("Advanced Settings", open=False):
+#                     seed = gr.Slider(
+#                         label="Seed",
+#                         minimum=0,
+#                         maximum=MAX_SEED,
+#                         step=1,
+#                         value=0,
+#                     )
+
+#                     randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+
+#                     with gr.Row():
+#                         guidance_scale = gr.Slider(
+#                             label="Guidance scale",
+#                             minimum=0.0,
+#                             maximum=10.0,
+#                             step=0.1,
+#                             value=7.0,  # Replace with defaults that work for your model
+#                         )
+#                     with gr.Row():
+#                         num_inference_steps = gr.Slider(
+#                             label="Number of inference steps - 50 inference steps are recommended; but you can reduce to 20 if the demo fails.",
+#                             minimum=1,
+#                             maximum=50,
+#                             step=1,
+#                             value=25,  # Replace with defaults that work for your model
+#                         )
+#                     with gr.Row():
+#                         num_of_interpolation = gr.Slider(
+#                             label="Number of images for interpolation - More images yield smoother transitions but require more resources and may fail.",
+#                             minimum=5,
+#                             maximum=50,
+#                             step=1,
+#                             value=10,  # Replace with defaults that work for your model
+#                         )
+
+#                 gr.Examples(examples=examples, inputs=[prompt1, prompt2])
+
+#             with gr.Tab("Arithmetic Operations"):
+#                 # The second tab is currently empty. You can add more components later.
+#                 gr.Markdown("This demo only supports addition or subtraction between two text latents ('Prompt_1 + Prompt_2' or 'Prompt_1 - Prompt_2'). For the other arithmetic operations, see the original [code](https://github.com/qihao067/CrossFlow).")
+#                 with gr.Row():
+#                     prompt1 = gr.Text(
+#                         label="Prompt_1",
+#                         show_label=False,
+#                         max_lines=1,
+#                         placeholder="Enter your prompt for the first image",
+#                         container=False,
+#                     )
                 
-    #             with gr.Row():
-    #                 prompt2 = gr.Text(
-    #                     label="Prompt_2",
-    #                     show_label=False,
-    #                     max_lines=1,
-    #                     placeholder="Enter your prompt for the second image",
-    #                     container=False,
-    #                 )
-
-    #             with gr.Row():
-    #                 operation_mode = gr.Radio(
-    #                     choices=["Addition", "Subtraction"],
-    #                     label="Operation Mode",
-    #                     value="Addition",
-    #                 )
-    #             with gr.Row():
-    #                 run_button = gr.Button("Run", scale=0, variant="primary")
-
-    #             # Create separate outputs for the first image, last image, and the animated GIF
-    #             first_image_output = gr.Image(label="Image of the first prompt", show_label=True)
-    #             last_image_output = gr.Image(label="Image of the second prompt", show_label=True)
-    #             gif_output = gr.Image(label="Linear interpolation", show_label=True)
-
-    #             with gr.Accordion("Advanced Settings", open=False):
-    #                 seed = gr.Slider(
-    #                     label="Seed",
-    #                     minimum=0,
-    #                     maximum=MAX_SEED,
-    #                     step=1,
-    #                     value=0,
-    #                 )
-
-    #                 randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-
-    #                 with gr.Row():
-    #                     guidance_scale = gr.Slider(
-    #                         label="Guidance scale",
-    #                         minimum=0.0,
-    #                         maximum=10.0,
-    #                         step=0.1,
-    #                         value=7.0,  # Replace with defaults that work for your model
-    #                     )
-    #                 # with gr.Row():
-    #                 #     num_inference_steps = gr.Slider(
-    #                 #         label="Number of inference steps - 50 inference steps are recommended; but you can reduce to 20 if the demo fails.",
-    #                 #         minimum=1,
-    #                 #         maximum=50,
-    #                 #         step=1,
-    #                 #         value=55,  # Replace with defaults that work for your model
-    #                 #     )
-
-    #                 with gr.Row():
-    #                     num_of_interpolation = gr.Slider(
-    #                         label="Number of images for interpolation - More images yield smoother transitions but require more resources and may fail.",
-    #                         minimum=5,
-    #                         maximum=50,
-    #                         step=1,
-    #                         value=50,  # Replace with defaults that work for your model
-    #                     )
-
-    #             gr.Examples(examples=examples, inputs=[prompt1, prompt2])
-
-    # gr.on(
-    #     triggers=[run_button.click, prompt1.submit, prompt2.submit],
-    #     fn=infer,
-    #     inputs=[
-    #         prompt1,
-    #         prompt2,
-    #         seed,
-    #         randomize_seed,
-    #         guidance_scale,
-    #         num_inference_steps,
-    #         num_of_interpolation,
-    #     ],
-    #     outputs=[first_image_output, last_image_output, gif_output, seed],
-    #     # outputs=[first_image_output, last_image_output, seed],
-    # )
+#                 with gr.Row():
+#                     prompt2 = gr.Text(
+#                         label="Prompt_2",
+#                         show_label=False,
+#                         max_lines=1,
+#                         placeholder="Enter your prompt for the second image",
+#                         container=False,
+#                     )
+
+#                 with gr.Row():
+#                     operation_mode = gr.Radio(
+#                         choices=["Addition", "Subtraction"],
+#                         label="Operation Mode",
+#                         value="Addition",
+#                     )
+#                 with gr.Row():
+#                     run_button = gr.Button("Run", scale=0, variant="primary")
+
+#                 # Create separate outputs for the first image, last image, and the animated GIF
+#                 first_image_output = gr.Image(label="Image of the first prompt", show_label=True)
+#                 last_image_output = gr.Image(label="Image of the second prompt", show_label=True)
+#                 gif_output = gr.Image(label="Linear interpolation", show_label=True)
+
+#                 with gr.Accordion("Advanced Settings", open=False):
+#                     seed = gr.Slider(
+#                         label="Seed",
+#                         minimum=0,
+#                         maximum=MAX_SEED,
+#                         step=1,
+#                         value=0,
+#                     )
+
+#                     randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
+
+#                     with gr.Row():
+#                         guidance_scale = gr.Slider(
+#                             label="Guidance scale",
+#                             minimum=0.0,
+#                             maximum=10.0,
+#                             step=0.1,
+#                             value=7.0,  # Replace with defaults that work for your model
+#                         )
+#                     # with gr.Row():
+#                     #     num_inference_steps = gr.Slider(
+#                     #         label="Number of inference steps - 50 inference steps are recommended; but you can reduce to 20 if the demo fails.",
+#                     #         minimum=1,
+#                     #         maximum=50,
+#                     #         step=1,
+#                     #         value=55,  # Replace with defaults that work for your model
+#                     #     )
+
+#                     with gr.Row():
+#                         num_of_interpolation = gr.Slider(
+#                             label="Number of images for interpolation - More images yield smoother transitions but require more resources and may fail.",
+#                             minimum=5,
+#                             maximum=50,
+#                             step=1,
+#                             value=50,  # Replace with defaults that work for your model
+#                         )
+
+#                 gr.Examples(examples=examples, inputs=[prompt1, prompt2])
+
+#     gr.on(
+#         triggers=[run_button.click, prompt1.submit, prompt2.submit],
+#         fn=infer,
+#         inputs=[
+#             prompt1,
+#             prompt2,
+#             seed,
+#             randomize_seed,
+#             guidance_scale,
+#             num_inference_steps,
+#             num_of_interpolation,
+#         ],
+#         outputs=[first_image_output, last_image_output, gif_output, seed],
+#         # outputs=[first_image_output, last_image_output, seed],
+#     )
 
 with gr.Blocks(css=css) as demo:
     with gr.Tabs():
         # --- Tab 1: Interpolation Mode (no operation_mode) ---
-        with gr.Tab("Tab 1: Interpolation Mode"):
-            gr.Markdown("**Tab 1:** This tab uses a slider for the number of interpolated images. The operation mode is fixed to *Addition* by default.")
+        gr.Markdown("# CrossFlow")
+        gr.Markdown("[CrossFlow](https://cross-flow.github.io/) directly transforms text representations into images for text-to-image generation, without the need for both the noise distribution and conditioning mechanism.")
+        gr.Markdown("This direct mapping enables meaningful 'Linear Interpolation' and 'Arithmetic Operations' in the text latent space, as demonstrated here.")
+        with gr.Tab("##  Linear Interpolation"):
+            gr.Markdown("This demo uses 256px images, 25 sampling steps (instead of 50), and 10 interpolations (instead of 50) to conserve GPU memory. For better results, see the original [code](https://github.com/qihao067/CrossFlow). (You may adjust them in Advanced Settings, but doing so may trigger OOM errors.)")
             
             prompt1_tab1 = gr.Text(placeholder="Prompt for first image", label="Prompt 1")
             prompt2_tab1 = gr.Text(placeholder="Prompt for second image", label="Prompt 2")
@@ -485,21 +488,21 @@ with gr.Blocks(css=css) as demo:
             )
         
         # --- Tab 2: Operation Mode (no num_of_interpolation) ---
-        with gr.Tab("Tab 2: Operation Mode"):
-            gr.Markdown("**Tab 2:** This tab lets you choose the operation mode (Addition or Subtraction) while fixing the number of interpolations to 3.")
+        with gr.Tab("##  Arithmetic Operations"):
+            gr.Markdown("This demo only supports addition or subtraction between two text latents ('Prompt_1 + Prompt_2' or 'Prompt_1 - Prompt_2'). For the other arithmetic operations, see the original [code](https://github.com/qihao067/CrossFlow).")
             
             prompt1_tab2 = gr.Text(placeholder="Prompt for first image", label="Prompt 1")
             prompt2_tab2 = gr.Text(placeholder="Prompt for second image", label="Prompt 2")
             seed_tab2 = gr.Slider(minimum=0, maximum=MAX_SEED, step=1, value=0, label="Seed")
             randomize_seed_tab2 = gr.Checkbox(label="Randomize seed", value=True)
             guidance_scale_tab2 = gr.Slider(minimum=0.0, maximum=10.0, step=0.1, value=7.0, label="Guidance Scale")
-            num_inference_steps_tab2 = gr.Slider(minimum=1, maximum=50, step=1, value=25, label="Number of Inference Steps")
+            num_inference_steps_tab2 = gr.Slider(minimum=1, maximum=50, step=1, value=50, label="Number of Inference Steps")
             operation_mode_tab2 = gr.Radio(choices=["Addition", "Subtraction"], label="Operation Mode", value="Addition")
             run_button_tab2 = gr.Button("Run")
             
             first_image_output_tab2 = gr.Image(label="Image of the first prompt")
             last_image_output_tab2 = gr.Image(label="Image of the second prompt")
-            gif_output_tab2 = gr.Image(label="Linear interpolation")
+            gif_output_tab2 = gr.Image(label="Resulting image produced by the arithmetic operations.")
             
             run_button_tab2.click(
                 fn=infer_tab2,
@@ -516,6 +519,5 @@ with gr.Blocks(css=css) as demo:
             )
 
 
-
 if __name__ == "__main__":
     demo.launch()