3D variational autoencoder for fingerprinting microstructure volume elements: Supplementary Data

Supplementary data and model checkpoints for the code found at https://github.com/micmog/lightning-vae3d. This is supplementary material for the paper found at https://arxiv.org/abs/2503.17427. File structure: datasets.zip mlp_data (training and validation data for the fully connected surrogate model - consists of microstructural fingerprints, load paths and stress responses) vae_data (training and validation data used for training the variational autoencoder - stored as .pt files for loading as PyTorch tensors) d3d_3d_train_fz.zip d3d_3d_val_fz.zip checkpoints.zip mlp_checkpoints (contains a checkpoint for the trained fully connected network, which acts as a surrogate model for uniaxial crystal plasticity simulations) vae_checkpoints (contains checkpoints and metadata files for the VAE architecture)本数据集为https://github.com/micmog/lightning-vae3d仓库中配套代码的补充数据与模型检查点，同时也是arXiv论文https://arxiv.org/abs/2503.17427的补充材料。 文件结构如下： - datasets.zip - mlp_data：包含用于训练多层感知器（Multi-Layer Perceptron, MLP）全连接替代模型的训练与验证数据，涵盖微观结构指纹、载荷路径与应力响应数据 - vae_data：包含用于训练变分自编码器（Variational AutoEncoder）的训练与验证数据集，以PyTorch张量格式存储为.pt文件 - d3d_3d_train_fz.zip - d3d_3d_val_fz.zip - checkpoints.zip：其内部包含如下目录： - mlp_checkpoints：包含训练完成的全连接网络的模型检查点，该网络可作为单轴晶体塑性仿真的替代模型 - vae_checkpoints：包含变分自编码器架构的模型检查点与元数据文件