Technical University of Denmark
Browse
DiffEnc-8-2_mnist_train_noise.zip (764.67 MB)

DiffEnc-8-2 trained on MNIST with trainable noise schedule

Download (764.67 MB)
dataset
posted on 2024-04-22, 10:00 authored by Beatrix Miranda Ginn NielsenBeatrix Miranda Ginn Nielsen

Checkpoints for an image generation model trained on MNIST.

The model was made in Jax. See the github repository for code to load the checkpoints.

The model is a variational diffusion model (VDM, https://arxiv.org/abs/2107.00630) with an added trainable time-dependent encoder trained for the article "DiffEnc: Variational Diffusion with a Learned Encoder" (https://arxiv.org/abs/2310.19789).

Model uses v-parametrization for the loss. The diffusion model is of size 8 and the encoder is of size 2. That is, the diffusion model uses 8 "down-blocks" in the U-net. See details in article.

Model was trained on MNIST for 2 million steps.

Random seeds: 1, 2, 13, 42, 70

Funding

Danish Pioneer Centre for AI, DNRF grant number P1

History

ORCID for corresponding depositor

Usage metrics

    DTU Compute

    Categories

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC