up

README.md

@@ -1,15 +1,22 @@
 ---
 tags:
-- ddpm_diffusion
+- ddim_diffusion
 ---
 
-# Denoising Diffusion Probabilistic Models (DDPM)
+# Denoising Diffusion Implicit Models (DDIM)
 
-**Paper**: [Denoising Diffusion Probabilistic Models](https://arxiv.org/abs/2006.11239)
+**Paper**: [Denoising Diffusion Implicit Models](https://arxiv.org/abs/2010.02502)
 
 **Abstract**:
 
-*We present high quality image synthesis results using diffusion probabilistic models, a class of latent variable models inspired by considerations from nonequilibrium thermodynamics. Our best results are obtained by training on a weighted variational bound designed according to a novel connection between diffusion probabilistic models and denoising score matching with Langevin dynamics, and our models naturally admit a progressive lossy decompression scheme that can be interpreted as a generalization of autoregressive decoding. On the unconditional CIFAR10 dataset, we obtain an Inception score of 9.46 and a state-of-the-art FID score of 3.17. On 256x256 LSUN, we obtain sample quality similar to ProgressiveGAN.*
+*Denoising diffusion probabilistic models (DDPMs) have achieved high quality image generation without adversarial training, yet they require simulating a Markov chain for many steps to produce a sample. To accelerate sampling, we present denoising diffusion implicit models (DDIMs), a more efficient class of iterative implicit probabilistic models with the same training procedure as DDPMs. In DDPMs, the generative process is defined as the reverse of a Markovian diffusion process. We construct a class of non-Markovian diffusion processes that lead to the same training objective, but whose reverse process can be much faster to sample from. We empirically demonstrate that DDIMs can produce high quality samples 10× to 50× faster in terms of wall-clock time compared to DDPMs, allow us to trade off computation for sample quality, and can perform semantically meaningful image interpolation directly in the latent space.*
+
+**Explanation on `eta` and `num_inference_steps`**
+
+- `num_inference_steps` is called *S* in the following table
+- `eta` is called *η* in the following table
+
+![ddim](https://huggingface.co/datasets/patrickvonplaten/scientific_images/resolve/main/table_ddim.png)
 
 ## Usage
 
@@ -19,13 +26,13 @@ from diffusers import DiffusionPipeline
 import PIL.Image
 import numpy as np
 
-model_id = "fusing/ddpm-celeba-hq"
+model_id = "fusing/ddim-celeba-hq"
 
 # load model and scheduler
 ddpm = DiffusionPipeline.from_pretrained(model_id)
 
 # run pipeline in inference (sample random noise and denoise)
-image = ddpm()
+image = ddpm(eta=0.0, num_inference_steps=50)
 
 # process image to PIL
 image_processed = image.cpu().permute(0, 2, 3, 1)
@@ -39,7 +46,7 @@ image_pil.save("test.png")
 
 ## Samples
 
-1. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddpm-celeba-hq/image_0.png)
-2. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddpm-celeba-hq/image_1.png)
-3. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddpm-celeba-hq/image_2.png)
-4. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddpm-celeba-hq/image_3.png)
+1. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddim-celeba-hq/image_0.png)
+2. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddim-celeba-hq/image_1.png)
+3. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddim-celeba-hq/image_2.png)
+4. ![sample_1](https://huggingface.co/datasets/patrickvonplaten/images/resolve/main/hf/ddim-celeba-hq/image_3.png)