DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation

TL;DR

DiffusionBlocks introduces a theoretically grounded block-wise training framework for transformer networks, reducing memory bottlenecks while maintaining performance across diverse architectures and tasks.

Contribution

It transforms transformer networks into independent trainable blocks using a diffusion interpretation, enabling scalable training with minimal modifications.

Findings

Matches end-to-end training performance

Reduces memory requirements proportionally to number of blocks

Effective across various transformer architectures

Abstract

End-to-end backpropagation requires storing activations throughout all layers, creating memory bottlenecks that limit model scalability. Existing block-wise training methods offer means to alleviate this problem, but they rely on ad-hoc local objectives and remain largely unexplored beyond classification tasks. We propose $\textit{DiffusionBlocks}$, a principled framework for transforming transformer-based networks into genuinely independent trainable blocks that maintain competitive performance with end-to-end training. Our key insight leverages the fact that residual connections naturally correspond to updates in a dynamical system. With minimal modifications to this system, we can convert the updates to those of a denoising process, where each block can be learned independently by leveraging the score matching objective. This independence enables training with gradients for only one block at a time, thereby reducing memory requirements in proportion to the number of blocks. Our experiments on a range of transformer architectures (vision, diffusion, autoregressive, recurrent-depth, and masked diffusion) demonstrate that DiffusionBlocks training matches the performance of end-to-end training while enabling scalable block-wise training on practical tasks beyond small-scale classification. DiffusionBlocks provides a theoretically grounded approach that successfully scales to modern generative tasks across diverse architectures. Code is available at https://github.com/SakanaAI/DiffusionBlocks .