Debug cac support

README.md

@@ -7,7 +7,7 @@ sdk: gradio
 sdk_version: 3.9
 app_file: app.py
 pinned: false
-license: mit
+license: apache-2.0
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -228,7 +228,7 @@ def get_equalizer(text: str, word_select: Union[int, Tuple[int, ...]], values: U
 
 def inference(source_prompt, target_prompt, source_guidance_scale=1, guidance_scale=5, num_inference_steps=100,
               width=512, height=512, seed=0, img=None, strength=0.7,
-              cross_attention_control=None, cross_replace_steps=0.8, self_replace_steps=0.4):
+              cross_attention_control="None", cross_replace_steps=0.8, self_replace_steps=0.4):
 
     torch.manual_seed(seed)
 
@@ -236,6 +236,7 @@ def inference(source_prompt, target_prompt, source_guidance_scale=1, guidance_sc
     img = img.resize((int(img.width * ratio), int(img.height * ratio)))
 
     # create the CAC controller.
+    assert cross_attention_control in ['Replace', 'Refine', "None"]
     if cross_attention_control == "replace":
         controller = AttentionReplace([source_prompt, target_prompt],
                                       num_inference_steps,
@@ -282,8 +283,8 @@ with gr.Blocks(css=css) as demo:
               </div>
               <p>
                 Demo for CycleDiffusion with Stable Diffusion. <br>
-                CycleDiffusion (<a href="https://github.com/ChenWu98/cycle-diffusion">Github</a> | <a href="https://arxiv.org/abs/2210.05559">📄 Paper link</a> | <a href="https://huggingface.co/docs/diffusers/main/en/api/pipelines/cycle_diffusion">🧨 Pipeline doc</a>) is an image-to-image translation method that supports stochastic samplers for diffusion models. <br>
-                It also supports Cross Attention Control (<a href="https://github.com/google/prompt-to-prompt">Github</a> | <a href="https://arxiv.org/abs/2208.01626">📄 Paper link</a>), which is a technique to transfer the attention map from the source prompt to the target prompt. <br>
+                CycleDiffusion (<a href="https://arxiv.org/abs/2210.05559">📄 Paper link</a> | <a href="https://huggingface.co/docs/diffusers/main/en/api/pipelines/cycle_diffusion">🧨 Pipeline doc</a>) is an image-to-image translation method that supports stochastic samplers for diffusion models. <br>
+                It also supports Cross Attention Control (<a href="https://arxiv.org/abs/2208.01626">📄 Paper link</a>), which is a technique to transfer the attention map from the source prompt to the target prompt. <br>
               </p>
               <p>You can skip the queue in the colab: <a href="https://colab.research.google.com/gist/ChenWu98/0aa4fe7be80f6b45d3d055df9f14353a/copy-of-fine-tuned-diffusion-gradio.ipynb"><img data-canonical-src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab" src="https://colab.research.google.com/assets/colab-badge.svg"></a></p>
                Running on <b>{device_print}</b>{(" in a <b>Google Colab</b>." if is_colab else "")}
@@ -314,9 +315,8 @@ with gr.Blocks(css=css) as demo:
                         guidance_scale = gr.Slider(label="Target guidance scale", value=5, minimum=1, maximum=10)
                     with gr.Row():
                         strength = gr.Slider(label="Strength", value=0.7, minimum=0.5, maximum=1, step=0.01)
-
                     with gr.Row():
-                        generate = gr.Button(value="Edit")
+                        generate1 = gr.Button(value="Edit")
             with gr.Tab("Basic options"):
                 with gr.Group():
                     with gr.Row():
@@ -326,6 +326,8 @@ with gr.Blocks(css=css) as demo:
 
                     with gr.Row():
                         seed = gr.Slider(0, 2147483647, label='Seed', value=0, step=1)
+                    with gr.Row():
+                        generate2 = gr.Button(value="Edit")
 
             with gr.Tab("CAC options"):
                 with gr.Group():
@@ -335,11 +337,15 @@ with gr.Blocks(css=css) as demo:
                         # If not "None", the following two parameters will be used.
                         cross_replace_steps = gr.Slider(label="Cross replace steps", value=0.8, minimum=0.0, maximum=1, step=0.01)
                         self_replace_steps = gr.Slider(label="Self replace steps", value=0.4, minimum=0.0, maximum=1, step=0.01)
+                    with gr.Row():
+                        generate3 = gr.Button(value="Edit")
 
     inputs = [source_prompt, target_prompt, source_guidance_scale, guidance_scale, num_inference_steps,
               width, height, seed, img, strength,
               cross_attention_control, cross_replace_steps, self_replace_steps]
-    generate.click(inference, inputs=inputs, outputs=image_out)
+    generate1.click(inference, inputs=inputs, outputs=image_out)
+    generate2.click(inference, inputs=inputs, outputs=image_out)
+    generate3.click(inference, inputs=inputs, outputs=image_out)
 
     ex = gr.Examples(
         [