Update README.md

README.md

@@ -20,10 +20,28 @@ tags:
 
 ## Usage
 
-### Unrolled loop
+### Inference with a pipeline
 
 ```python
-!pip install git+https://github.com/huggingface/diffusers.git
+!pip install diffusers
+from diffusers import DiffusionPipeline
+
+model_id = "CompVis/ldm-celebahq-256"
+
+# load model and scheduler
+pipeline = DiffusionPipeline.from_pretrained(model_id)
+
+# run pipeline in inference (sample random noise and denoise)
+image = pipeline(num_inference_steps=200)["sample"]
+
+# save image
+image[0].save("ldm_generated_image.png")
+```
+
+### Inference with an unrolled loop
+
+```python
+!pip install diffusers
 from diffusers import UNet2DModel, DDIMScheduler, VQModel
 import torch
 import PIL.Image
@@ -33,9 +51,9 @@ import tqdm
 seed = 3
 
 # load all models
-unet = UNet2DModel.from_pretrained("CompVis/latent-diffusion-celeba-256", subfolder="unet")
-vqvae = VQModel.from_pretrained("CompVis/latent-diffusion-celeba-256", subfolder="vqvae")
-scheduler = DDIMScheduler.from_config("CompVis/latent-diffusion-celeba-256", subfolder="scheduler")
+unet = UNet2DModel.from_pretrained("CompVis/ldm-celebahq-256", subfolder="unet")
+vqvae = VQModel.from_pretrained("CompVis/ldm-celebahq-256", subfolder="vqvae")
+scheduler = DDIMScheduler.from_config("CompVis/ldm-celebahq-256", subfolder="scheduler")
 
 # set to cuda
 torch_device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -46,7 +64,7 @@ vqvae.to(torch_device)
 # generate gaussian noise to be decoded
 generator = torch.manual_seed(seed)
 noise = torch.randn(
-    (1, unet.in_channels, unet.image_size, unet.image_size),
+    (1, unet.in_channels, unet.sample_size, unet.sample_size),
     generator=generator,
 ).to(torch_device)
 
@@ -78,36 +96,10 @@ image_pil = PIL.Image.fromarray(image_processed[0])
 image_pil.save(f"generated_image_{seed}.png")
 ```
 
-### pipeline
-
-```python
-!pip install git+https://github.com/huggingface/diffusers.git
-from diffusers import LatentDiffusionUncondPipeline
-import torch
-import PIL.Image
-import numpy as np
-import tqdm
-
-seed = 3
-
-pipeline = LatentDiffusionUncondPipeline.from_pretrained("CompVis/latent-diffusion-celeba-256")
-
-# generatae image by calling the pipeline
-generator = torch.manual_seed(seed)
-image = pipeline(generator=generator, num_inference_steps=200)["sample"]
-
-# process image
-image_processed = image.cpu().permute(0, 2, 3, 1)
-image_processed = (image_processed + 1.0) * 127.5
-image_processed = image_processed.clamp(0, 255).numpy().astype(np.uint8)
-image_pil = PIL.Image.fromarray(image_processed[0])
-
-image_pil.save(f"generated_image_{seed}.png")
-```
 
 ## Samples
 
-1. ![sample_1](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/generated_image_0.png)
-2. ![sample_1](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/generated_image_1.png)
-3. ![sample_1](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/generated_image_2.png)
-4. ![sample_1](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/generated_image_3.png)
+1. ![sample_0](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/images/generated_image_0.png)
+2. ![sample_1](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/images/generated_image_1.png)
+3. ![sample_2](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/images/generated_image_2.png)
+4. ![sample_3](https://huggingface.co/CompVis/latent-diffusion-celeba-256/resolve/main/images/generated_image_3.png)