Structural Pruning for Diffusion Models

Authors: Gongfan Fang, Xinyin Ma, Xinchao Wang

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical assessment, undertaken across several datasets highlights two primary benefits of our proposed method: 1) Efficiency: it enables approximately a 50% reduction in FLOPs at a mere 10% to 20% of the original training expenditure; 2) Consistency: the pruned diffusion models inherently preserve generative behavior congruent with their pre-trained models. Code is available at https://github.com/Vain F/Diff-Pruning.
Researcher Affiliation Academia Gongfan Fang Xinyin Ma Xinchao Wang National University of Singapore gongfan@u.nus.edu, maxinyin@u.nus.edu, xinchao@nus.edu.sg
Pseudocode Yes Algorithm 1 Diff-Pruning
Open Source Code Yes Code is available at https://github.com/Vain F/Diff-Pruning.
Open Datasets Yes The efficacy of Diff-Pruning is empirically validated across six diverse datasets, including CIFAR-10 (32 × 32)[21], Celeb A-HQ (64 × 64)[32], LSUN Church (256 × 256), LSUN Bedroom (256 × 256) [57] and Image Net-1K (256 × 256).
Dataset Splits No The paper uses well-known public datasets, but it does not explicitly state the dataset splits (e.g., percentages, sample counts) for training, validation, or testing within the text.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types) used for the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers.
Experiment Setup Yes For example, when applying our method to an off-the-shelf diffusion model pre-trained on LSUN Church [57], we achieve an impressive compression rate of 50% FLOPs, with only 10% of the training cost required by the original models, equating to 0.5 million steps compared to the 4.4 million steps of the pre-existing models.