Update app.py

app.py

@@ -31,11 +31,9 @@ import nvidia_smi
 nvidia_smi.nvmlInit()
 handle = nvidia_smi.nvmlDeviceGetHandleByIndex(0)
 # card id 0 hardcoded here, there is also a call to get all available card ids, so we could iterate
-
 torch.hub.download_url_to_file('https://images.pexels.com/photos/158028/bellingrath-gardens-alabama-landscape-scenic-158028.jpeg', 'garden.jpeg')
 torch.hub.download_url_to_file('https://images.pexels.com/photos/68767/divers-underwater-ocean-swim-68767.jpeg', 'coralreef.jpeg')
 torch.hub.download_url_to_file('https://images.pexels.com/photos/803975/pexels-photo-803975.jpeg', 'cabin.jpeg')
-
 def sinc(x):
     return torch.where(x != 0, torch.sin(math.pi * x) / (math.pi * x), x.new_ones([]))
 def lanczos(x, a):
@@ -200,6 +198,7 @@ device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
 print('Using device:', device)
 model = load_vqgan_model(args.vqgan_config, args.vqgan_checkpoint).to(device)
 perceptor = clip.load(args.clip_model, jit=False)[0].eval().requires_grad_(False).to(device)
+all_frames = []
 def inference(text, seed, step_size, max_iterations, width, height, init_image, init_weight):
     size=[width, height]
     texts = text
@@ -319,7 +318,9 @@ def inference(text, seed, step_size, max_iterations, width, height, init_image,
             result.append(prompt(iii))
         img = np.array(out.mul(255).clamp(0, 255)[0].cpu().detach().numpy().astype(np.uint8))[:,:,:]
         img = np.transpose(img, (1, 2, 0))
-        #imageio.imwrite('./steps/' + str(i) + '.png', np.array(img))
+        # imageio.imwrite('./steps/' + str(i) + '.png', np.array(img))
+        img = Image.fromarray(img).convert('RGB')
+        all_frames.append(img)
         return result, np.array(img)
     def train(i):
         opt.zero_grad()
@@ -344,13 +345,14 @@ def inference(text, seed, step_size, max_iterations, width, height, init_image,
                 pbar.update()
     except KeyboardInterrupt:
         pass
-    return image
+    all_frames[0].save('out.gif',
+               save_all=True, append_images=all_frames[1:], optimize=False, duration=40, loop=0)
+    return image, 'out.gif'
 def load_image( infilename ) :
     img = Image.open( infilename )
     img.load()
     data = np.asarray( img, dtype="int32" )
     return data
-
 title = "VQGAN + CLIP"
 description = "Gradio demo for VQGAN + CLIP. To use it, simply add your text, or click one of the examples to load them. Read more at the links below. Please click submit only once. Results will show up in under a minute."
 article = "<p style='text-align: center'>Originally made by Katherine Crowson (https://github.com/crowsonkb, https://twitter.com/RiversHaveWings). The original BigGAN+CLIP method was by https://twitter.com/advadnoun. Added some explanations and modifications by Eleiber#8347, pooling trick by Crimeacs#8222 (https://twitter.com/EarthML1) and the GUI was made with the help of Abulafia#3734. | <a href='https://colab.research.google.com/drive/1ZAus_gn2RhTZWzOWUpPERNC0Q8OhZRTZ'>Colab</a> | <a href='https://github.com/CompVis/taming-transformers'>Taming Transformers Github Repo</a> | <a href='https://github.com/openai/CLIP'>CLIP Github Repo</a> | Special thanks to BoneAmputee (https://twitter.com/BoneAmputee) for suggestions and advice</p>"
@@ -365,7 +367,7 @@ gr.Interface(
     gr.inputs.Image(type="file", label="Initial Image"),
     gr.inputs.Slider(minimum=0.0, maximum=15.0, default=0.0, label='Initial Weight', step=1.0),
      ], 
-    gr.outputs.Image(type="numpy", label="Output"),
+    [gr.outputs.Image(type="numpy", label="Output Image"),gr.outputs.Image(type="file", label="Output GIF")],
     title=title,
     description=description,
     article=article,