Update app.py

app.py

@@ -22,12 +22,16 @@ h1 {
 }
 """
 
-def infer(prompt, image_inp, seed_inp, sampling_steps,width,height):
+def infer(prompt, image_inp, seed_inp, sampling_steps,width,height,infer_type):
     setup_seed(seed_inp)
     args.num_sampling_steps = sampling_steps
     img = cv2.imread(image_inp)
     new_size = [height,width]  
     args.image_size = new_size
+    if infer_type == 'ddpm':
+        args.sample_method = 'ddpm'
+    elif infer_type == 'ddim':
+        args.sample_method = 'ddim'
     vae, model, text_encoder, diffusion = model_i2v_fun(args)
     vae.to(device)
     model.to(device)
@@ -106,8 +110,10 @@ with gr.Blocks(css='style.css') as demo:
                 
                 with gr.Row():
                     # control_task = gr.Dropdown(label="Task", choices=["Text-2-video", "Image-2-video"], value="Text-2-video", multiselect=False, elem_id="controltask-in")
-                    sampling_steps = gr.Slider(label='Steps', minimum=50, maximum=300, value=250, step=1)
+                    sampling_steps = gr.Slider(label='Steps', minimum=50, maximum=300, value=100, step=1)
                     seed_inp = gr.Slider(label="Seed", minimum=0, maximum=2147483647, step=1, value=250, elem_id="seed-in")
+                with gr.Row():
+                    infer_type = gr.Dropdown(['ddpm','ddim'], label='infer_type',value='ddim')
                 with gr.Row():
                     width = gr.Slider(label='width',minimum=1,maximum=2000,value=512,step=1)
                     height = gr.Slider(label='height',minimum=1,maximum=2000,value=320,step=1)
@@ -119,14 +125,14 @@ with gr.Blocks(css='style.css') as demo:
                 # clean_btn = gr.Button("Clean video")
 
         video_out = gr.Video(label="Video result", elem_id="video-output", width = 800)
-        inputs = [prompt,image_inp, seed_inp, sampling_steps,width,height]
+        inputs = [prompt,image_inp, seed_inp, sampling_steps,width,height,infer_type]
         outputs = [video_out]
         ex = gr.Examples(
-            examples = [["./input/i2v/The_picture_shows_the_beauty_of_the_sea_and_at_the_same.png","A video of the beauty of the sea",123,250,560,240],
-                        ["./input/i2v/The_picture_shows_the_beauty_of_the_sea.png","A video of the beauty of the sea",14717,250,560,240],
-                        ["./input/i2v/Close-up_essence_is_poured_from_bottleKodak_Vision.png","A video of close-up essence is poured from bottleKodak Vision",178135313,250,560,240]],
+            examples = [["./input/i2v/The_picture_shows_the_beauty_of_the_sea_and_at_the_same.png","A video of the beauty of the sea",123,250,560,240,'ddim'],
+                        ["./input/i2v/The_picture_shows_the_beauty_of_the_sea.png","A video of the beauty of the sea",14717,250,560,240,'ddim'],
+                        ["./input/i2v/Close-up_essence_is_poured_from_bottleKodak_Vision.png","A video of close-up essence is poured from bottleKodak Vision",178135313,250,560,240,'ddim']],
             fn = infer,
-            inputs = [image_inp, prompt, seed_inp, sampling_steps,width,height],
+            inputs = [image_inp, prompt, seed_inp, sampling_steps,width,height,infer_type],
             outputs=[video_out],
             cache_examples=False
 
@@ -134,7 +140,6 @@ with gr.Blocks(css='style.css') as demo:
         )
         ex.dataset.headers = [""]
        
-    # control_task.change(change_task_options, inputs=[control_task], outputs=[canny_opt, hough_opt, normal_opt], queue=False)
     # clean_btn.click(clean, inputs=[], outputs=[video_out], queue=False)
     submit_btn.click(infer, inputs, outputs)
     # share_button.click(None, [], [], _js=share_js)