Update gradio_app.py

gradio_app.py

@@ -142,7 +142,7 @@ def get_image(image1, prompt, image2, dim_steps=50, ddim_eta=1., fs=None, seed=1
 
     image1 = torch.from_numpy(image1).permute(2, 0, 1).float().cuda()
     input_h, input_w = image1.shape[1:]
-    image1 = (image_tensor1 / 255. - 0.5) * 2
+    image1 = (image1 / 255. - 0.5) * 2
 
     image2 = torch.from_numpy(image2).permute(2, 0, 1).float().cuda()
     input_h, input_w = image2.shape[1:]
@@ -153,15 +153,15 @@ def get_image(image1, prompt, image2, dim_steps=50, ddim_eta=1., fs=None, seed=1
     image_tensor1 = transform(image1).unsqueeze(1) # [c,1,h,w]
     # image2 = Image.open(file_list[2*idx+1]).convert('RGB')
     image_tensor2 = transform(image2).unsqueeze(1) # [c,1,h,w]
-    frame_tensor1 = repeat(image_tensor1, 'c t h w -> c (repeat t) h w', repeat=15)
-    frame_tensor2 = repeat(image_tensor2, 'c t h w -> c (repeat t) h w', repeat=1)
+    frame_tensor1 = repeat(image_tensor1, 'c t h w -> c (repeat t) h w', repeat=8)
+    frame_tensor2 = repeat(image_tensor2, 'c t h w -> c (repeat t) h w', repeat=8)
     videos = torch.cat([frame_tensor1, frame_tensor2], dim=1)
     # frame_tensor = torch.cat([frame_tensor1, frame_tensor1], dim=1)
     # _, filename = os.path.split(file_list[idx*2])
     
     ddim_sampler = DDIMSampler(model) if not multiple_cond_cfg else DDIMSampler_multicond(model)
     batch_size = noise_shape[0]
-    fs = torch.tensor([fs] * batch_size, dtype=torch.long, device=model.device)
+    fs = torch.tensor([fs], dtype=torch.long, device=model.device)
 
     if not text_input:
         prompts = [""]*batch_size