AdjointDPM: Adjoint Sensitivity Method for Gradient Backpropagation of Diffusion Probabilistic Models

by   Jiachun Pan, et al.

Existing customization methods require access to multiple reference examples to align pre-trained diffusion probabilistic models (DPMs) with user-provided concepts. This paper aims to address the challenge of DPM customization when the only available supervision is a differentiable metric defined on the generated contents. Since the sampling procedure of DPMs involves recursive calls to the denoising UNet, naïve gradient backpropagation requires storing the intermediate states of all iterations, resulting in extremely high memory consumption. To overcome this issue, we propose a novel method AdjointDPM, which first generates new samples from diffusion models by solving the corresponding probability-flow ODEs. It then uses the adjoint sensitivity method to backpropagate the gradients of the loss to the models' parameters (including conditioning signals, network weights, and initial noises) by solving another augmented ODE. To reduce numerical errors in both the forward generation and gradient backpropagation processes, we further reparameterize the probability-flow ODE and augmented ODE as simple non-stiff ODEs using exponential integration. Finally, we demonstrate the effectiveness of AdjointDPM on three interesting tasks: converting visual effects into identification text embeddings, finetuning DPMs for specific types of stylization, and optimizing initial noise to generate adversarial samples for security auditing.


page 8

page 9

page 14

page 15

page 16

page 17

page 18

page 19


Diffusion with Forward Models: Solving Stochastic Inverse Problems Without Direct Supervision

Denoising diffusion models are a powerful type of generative models used...

Conditional Generation from Unconditional Diffusion Models using Denoiser Representations

Denoising diffusion models have gained popularity as a generative modeli...

Pseudo Numerical Methods for Diffusion Models on Manifolds

Denoising Diffusion Probabilistic Models (DDPMs) can generate high-quali...

TRACT: Denoising Diffusion Models with Transitive Closure Time-Distillation

Denoising Diffusion models have demonstrated their proficiency for gener...

Improving Handwritten OCR with Training Samples Generated by Glyph Conditional Denoising Diffusion Probabilistic Model

Constructing a highly accurate handwritten OCR system requires large amo...

Memorized Sparse Backpropagation

Neural network learning is typically slow since backpropagation needs to...

A Graph Theoretic Framework of Recomputation Algorithms for Memory-Efficient Backpropagation

Recomputation algorithms collectively refer to a family of methods that ...

Please sign up or login with your details

Forgot password? Click here to reset