DiffEnc-32-4 trained on CIFAR-10

Checkpoints for an image generation model trained on CIFAR-10.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 32 and the encoder is of size 4. That is, the diffusion model uses 32 "down-blocks" in the U-net. See details in article.Model was trained on CIFAR-10 for 8 million steps.Random seeds: 1, 2, 13

本数据集为基于CIFAR-10训练的图像生成模型的检查点（checkpoint）文件。该模型基于Jax框架开发，加载该检查点的代码可参阅对应GitHub仓库。本模型为变分扩散模型（variational diffusion model, VDM，https://arxiv.org/abs/2107.00630），并额外搭载了可训练的时序编码器，源自论文《DiffEnc: Variational Diffusion with a Learned Encoder》（https://arxiv.org/abs/2310.19789）。模型训练时采用v参数化方式计算损失。该扩散模型的U-Net包含32个下采样块，编码器维度为4，详细细节可参阅上述论文。本模型在CIFAR-10数据集上共训练800万步，所用随机种子为1、2、13。