omerbartal commited on
Commit
ecbe199
1 Parent(s): 6e8765b

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +10 -10
app.py CHANGED
@@ -15,25 +15,25 @@ from diffusers import StableDiffusionPanoramaPipeline, DDIMScheduler
15
 
16
  model_ckpt = "stabilityai/stable-diffusion-2-base"
17
  scheduler = DDIMScheduler.from_pretrained(model_ckpt, subfolder="scheduler")
18
- # pipe = StableDiffusionPanoramaPipeline.from_pretrained(
19
- # model_ckpt, scheduler=scheduler, torch_dtype=torch.float16
20
- # )
21
  pipe = StableDiffusionPanoramaPipeline.from_pretrained(
22
- model_ckpt, scheduler=scheduler
23
  )
 
 
 
24
 
25
  pipe = pipe.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
26
 
27
- def generate_image_fn(prompt: str, guidance_scale: float) -> list:
28
  start_time = time.time()
29
  prompt = "a photo of the dolomites"
30
- image = pipe(prompt, guidance_scale=guidance_scale).images
31
  end_time = time.time()
32
  print(f"Time taken: {end_time - start_time} seconds.")
33
  return image
34
 
35
 
36
- description = "This Space demonstrates MultiDiffusion Text2Panorama using Stable Diffusion model. You can use it for generating custom pokemons. To get started, either enter a prompt and pick one from the examples below. For details on the fine-tuning procedure, refer to [this repository]()."
37
  article = "This Space leverages a T4 GPU to run the predictions. We use mixed-precision to speed up the inference latency."
38
  gr.Interface(
39
  generate_image_fn,
@@ -43,12 +43,12 @@ gr.Interface(
43
  max_lines=1,
44
  placeholder="a photo of the dolomites",
45
  ),
46
- gr.Slider(value=40, minimum=8, maximum=50, step=1),
47
  ],
48
  outputs=gr.Gallery().style(grid=[2], height="auto"),
49
- title="Generate custom pokemons",
50
  description=description,
51
  article=article,
52
- examples=[["a photo of the dolomites", 40]],
53
  allow_flagging=False,
54
  ).launch(enable_queue=True)
 
15
 
16
  model_ckpt = "stabilityai/stable-diffusion-2-base"
17
  scheduler = DDIMScheduler.from_pretrained(model_ckpt, subfolder="scheduler")
 
 
 
18
  pipe = StableDiffusionPanoramaPipeline.from_pretrained(
19
+ model_ckpt, scheduler=scheduler, torch_dtype=torch.float16
20
  )
21
+ # pipe = StableDiffusionPanoramaPipeline.from_pretrained(
22
+ # model_ckpt, scheduler=scheduler
23
+ # )
24
 
25
  pipe = pipe.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
26
 
27
+ def generate_image_fn(prompt: str, img_width: int, img_height=512) -> list:
28
  start_time = time.time()
29
  prompt = "a photo of the dolomites"
30
+ image = pipe(prompt, height=img_height, width=img_width).images
31
  end_time = time.time()
32
  print(f"Time taken: {end_time - start_time} seconds.")
33
  return image
34
 
35
 
36
+ description = "This Space demonstrates MultiDiffusion Text2Panorama using Stable Diffusion model. To get started, either enter a prompt and pick one from the examples below. For details, please visit [the project page](https://multidiffusion.github.io/)."
37
  article = "This Space leverages a T4 GPU to run the predictions. We use mixed-precision to speed up the inference latency."
38
  gr.Interface(
39
  generate_image_fn,
 
43
  max_lines=1,
44
  placeholder="a photo of the dolomites",
45
  ),
46
+ gr.Slider(value=4096, minimum=512, maximum=4608, step=128),
47
  ],
48
  outputs=gr.Gallery().style(grid=[2], height="auto"),
49
+ title="Generate a panoramic image!",
50
  description=description,
51
  article=article,
52
+ examples=[["a photo of the dolomites", 4096]],
53
  allow_flagging=False,
54
  ).launch(enable_queue=True)