Beyond the Training Domain: Robust Generative Transition State Models for Unseen Chemistry

Pretrained models as well as the curated datasets. AEFM: configs and database to train and run inference using adaptive equilibrium flow matching. OAReactDiff: codebase and training script to finetune the published OAReactDiff checkpoint with the new embedding layer, as well as with the newly developed self-supervised scheme. ReactOT: codebase and training/inference script to train and sample with ReactOT checkpoint_: Checkpoints for all models used in the work预训练模型（Pretrained Models）与精选数据集（curated datasets）。 AEFM：用于基于自适应均衡流匹配（adaptive equilibrium flow matching）开展训练与推理的配置文件及数据库。 OAReactDiff：用于结合新嵌入层与新研发的自监督方案（self-supervised scheme），对已发布的OAReactDiff模型检查点（checkpoint）进行微调的代码库与训练脚本。 ReactOT：用于基于ReactOT开展训练与采样的代码库及训练/推理脚本。 checkpoint_：本研究中所用全部模型的检查点文件（checkpoints）。