LoQI: Scalable Low-Energy Molecular Conformer Generation with Quantum Mechanical Accuracy

This dataset is the supplementary data/ folder from the root of the LoQI repository (conformer generation with diffusion and flow-matching models).<br>Repository: https://github.com/isayevlab/LoQIFolder location in project:data/ (project root)Contents of data/:chembl3d_stereo/processed/: preprocessed train/val/test tensors and metadata used by the codebaseloqi.ckpt: pretrained LoQI diffusion checkpointloqi_flow.ckpt: pretrained LoQI flow-matching checkpointsha256sums.txt: SHA-256 checksums for all files in this packageProcessed split files include:train_h.pt, val_h.pt, test_h.pt (processed molecule data)split-specific feature arrays/pickles (atom types, bond types, charges, aromatic/ring flags, hybridization, bond lengths, angles, dihedrals, valency, and related metadata)test-only helper files: test_rot_bonds_h.pt, test_cremp_h.pt, test_small_h.ptIntegrity verification:<br>From repository root, run:<br>sha256sums.txtAll entries should return “OK”.Notes:Paths in sha256sums.txt are repository-relative (starting with data/).If any file is regenerated, recompute checksums before redistribution.<br>

本数据集为LoQI仓库根目录下的补充`data/`文件夹，该仓库聚焦于基于扩散模型与流匹配模型的分子构象生成任务。
仓库地址：https://github.com/isayevlab/LoQI
项目内文件夹路径：`data/`（即项目根目录下）
`data/`文件夹内包含以下内容：
- `chembl3d_stereo/processed/`：存放经预处理后的训练、验证、测试张量与代码库所用的元数据
- `loqi.ckpt`：预训练LoQI扩散模型检查点
- `loqi_flow.ckpt`：预训练LoQI流匹配模型检查点
- `sha256sums.txt`：本数据包内所有文件的SHA-256校验和

预处理拆分文件包含：
`train_h.pt`、`val_h.pt`、`test_h.pt`（已处理的分子数据）
拆分专属的特征数组/Pickle格式序列化文件（涵盖原子类型、键类型、电荷、芳香性/环标记、杂化类型、键长、键角、二面角、价态及相关元数据）
仅用于测试的辅助文件：`test_rot_bonds_h.pt`、`test_cremp_h.pt`、`test_small_h.pt`

完整性验证：
在仓库根目录下执行`sha256sum -c sha256sums.txt`，所有条目均应返回“OK”。

注意事项：
`sha256sums.txt`内的路径为仓库相对路径（以`data/`开头）。
若重新生成任一文件，请在重新分发前重新计算校验和。