add step params

app_diffedit.py

@@ -29,6 +29,9 @@ def image_to_image(
     input_image: Image,
     source_prompt: str,
     target_prompt: str,
+    num_inference_steps: int,
+    start_step: int,
+    guidance_scale: float,
 ):
     run_task_time = 0
     time_cost_str = ''
@@ -39,17 +42,29 @@ def image_to_image(
         image=input_image,
         source_prompt=source_prompt,
         target_prompt=target_prompt,
+        num_inference_steps = num_inference_steps,
+        guidance_scale=guidance_scale,
     )
 
-    inv_latents = basepipeline.invert(prompt=source_prompt, image=input_image).latents
+    inv_latents = basepipeline.invert(
+        prompt=source_prompt,
+        image=input_image,
+        num_inference_steps = num_inference_steps,
+        guidance_scale=guidance_scale,
+    ).latents
+
+    # get inverse latents by start step
+    # inv_latents = inv_latents[-(start_step + 1)][None]
 
     output_image = basepipeline(
         prompt=target_prompt,
         mask_image=mask_image,
         image_latents=inv_latents,
         negative_prompt=source_prompt,
+        num_inference_steps = num_inference_steps,
+        guidance_scale=guidance_scale,
     ).images[0]
-    mask_image = Image.fromarray((mask_image.squeeze()*255).astype("uint8"), "L")
+    mask_image = Image.fromarray((mask_image.squeeze()*255).astype("uint8"), "L").resize(input_image.size, Image.LANCZOS)
 
     
     run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
@@ -74,10 +89,10 @@ def create_demo() -> gr.Blocks:
                 input_image_prompt = gr.Textbox(lines=1, label="Input Image Prompt", value=DEFAULT_SRC_PROMPT)
                 edit_prompt = gr.Textbox(lines=1, label="Edit Prompt", value=DEFAULT_EDIT_PROMPT)
             with gr.Column():
-                num_steps = gr.Slider(minimum=1, maximum=100, value=20, step=1, label="Num Steps")
+                num_inference_steps = gr.Slider(minimum=1, maximum=100, value=20, step=1, label="Num Inference Steps")
                 start_step = gr.Slider(minimum=1, maximum=100, value=15, step=1, label="Start Step")
             with gr.Column():
-                seed = gr.Number(label="Seed", value=8)
+                guidance_scale = gr.Slider(minimum=0, maximum=20, value=7.5, step=0.5, label="Guidance Scale")
                 g_btn = gr.Button("Edit Image")
                 
         with gr.Row():
@@ -91,7 +106,7 @@ def create_demo() -> gr.Blocks:
         
         g_btn.click(
             fn=image_to_image,
-            inputs=[input_image, input_image_prompt, edit_prompt],
+            inputs=[input_image, input_image_prompt, edit_prompt, num_inference_steps, start_step, guidance_scale],
             outputs=[output_image, mask_image, generated_cost],
         )