<feat> add diffusion inference steps control.

app.py

@@ -98,7 +98,7 @@ def init_basemodel():
 
 
 @spaces.GPU
-def process_image_and_text(condition_image, target_prompt, condition_image_prompt, task, random_seed, inpainting, fill_x1, fill_x2, fill_y1, fill_y2):
+def process_image_and_text(condition_image, target_prompt, condition_image_prompt, task, random_seed, num_steps, inpainting, fill_x1, fill_x2, fill_y1, fill_y2):
     # set up the model
     global pipe, current_task, transformer
     if current_task != task:
@@ -265,7 +265,7 @@ def process_image_and_text(condition_image, target_prompt, condition_image_promp
         height=512,
         width=512,
         num_frames=5,
-        num_inference_steps=50,
+        num_inference_steps=num_steps,
         guidance_scale=6.0,
         num_videos_per_prompt=1,
         generator=torch.Generator(device=pipe.transformer.device).manual_seed(random_seed),
@@ -315,8 +315,9 @@ def create_app():
                 )
                 gr.Markdown(notice, elem_id="notice")
                 target_prompt = gr.Textbox(lines=2, label="Target Prompt", elem_id="tp")
-                condition_image_prompt = gr.Textbox(lines=2, label="Condition Image Prompt", elem_id="cp")
+                condition_image_prompt = gr.Textbox(lines=2, label="Condition Image Prompt (Only required by Subject-driven Image Generation and Style Transfer tasks)", elem_id="cp")
                 random_seed = gr.Number(label="Random Seed", precision=0, value=0, elem_id="seed")
+                num_steps = gr.Number(label="Diffusion Inference Steps", precision=0, value=50, elem_id="steps")
                 inpainting = gr.Checkbox(label="Inpainting", value=False, elem_id="inpainting")
                 fill_x1 = gr.Number(label="In/Out-painting Box Left Boundary", precision=0, value=128, elem_id="fill_x1")
                 fill_x2 = gr.Number(label="In/Out-painting Box Right Boundary", precision=0, value=384, elem_id="fill_x2")
@@ -338,7 +339,7 @@ def create_app():
 
         submit_btn.click(
             fn=process_image_and_text,
-            inputs=[condition_image, target_prompt, condition_image_prompt, task, random_seed, inpainting, fill_x1, fill_x2, fill_y1, fill_y2],
+            inputs=[condition_image, target_prompt, condition_image_prompt, task, random_seed, num_steps, inpainting, fill_x1, fill_x2, fill_y1, fill_y2],
             outputs=output_images,
         )