Skip to content

README.md

Latest commit

 

History

History
78 lines (54 loc) · 3.1 KB

README.md

File metadata and controls

78 lines (54 loc) · 3.1 KB

OC20 and OC20-Dense Data Processing

This repository contains scripts and guidelines for handling the OC20 and OC20-Dense datasets, including steps for extracting relaxed positions, generating string representations, and integrating string files.

Original Datasets

The original datasets for OC20 and OC20-Dense can be accessed from the following links:

The preprocessed data and checkpoints can be accessed through the following link: Data. Find the pred_data and pred_checkpoint directory.

Note:

  • OC20 IS2RE LMDB files: Contain relaxed positions as pos_relaxed attribute.
  • OC20-Dense IS2RE LMDB files: Do not contain relaxed positions. To extract relaxed positions for the OC20-Dense dataset, an additional step is required.

LMDB Creation with Relaxed Frames (OC20-Dense)

We created a new LMDB using relaxed structures from publicly available trajectory files, and then sampled the training set from it. To create an LMDB file with the relaxed positions for the OC20-Dense dataset:

python relaxed_frame_lmdb.py --mapping_file_path <PATH_TO_MAPPING_FILE> \
                             --traj_path <DIRECTORY_OF_TRAJ_FILES> \
                             --save_path <PATH_TO_SAVE_FILE; lmdb extension> \
                             --refE_path <PATH_TO_REF_E_FILE> \
                             --tag_path <PATH_TO_TAG_FILE>

Required Files:

  • Mapping file: oc20dense_mapping.pkl
  • Reference Energy file: oc20dense_ref_energies.pkl
  • Tag file: oc20dense_tags.pkl

String Generation

After preparing the LMDB files, strings can be generated for further processing.

For OC20 Dataset:

python string_generator_oc20.py --lmdb_path <PATH_TO_LMDB> \
                                --save_path <PATH_TO_SAVE_FILE; pkl extension> \
                                --metadata_path <PATH_TO_METADATA> \
                                --task <is2re or rs2re or s2ef>

For OC20-Dense Dataset:

python string_generator_oc20dense.py --lmdb_path <PATH_TO_LMDB> \
                                     --save_path <PATH_TO_SAVE_FILE; pkl extension> \
                                     --metadata_path <PATH_TO_METADATA> \
                                     --split <train or val or eval or test>

Required Files:

  • OC20 metadata file: oc20_data_mapping.pkl
  • OC20-Dense metadata file: oc20dense_mapping.pkl

String File Integration

Once string files are generated, they can be integrated using the following command:

python string_integrator.py --string_dir <DIRECTORY_WHERE_STRINGS_SAVED> \
                            --save_path <PATH_TO_SAVE_FILE; pkl extension>