pretrained_models.md

Pretrained Models

Pretrained models are already released at Baidu Netdisk (Extract Code: lsgp) and Google Drive. After your download the pretrained models, put it in the main folder, then execute the following commands to unzip it.

unzip stats.zip

The stats folder should have the following structure.

stats
├── bert
|   └── checkpoint.tar
├── bert_cased
|   └── checkpoint.tar
├── citation_bert
|   └── checkpoint.tar
├── citation_bert_cased
|   └── checkpoint.tar
└── vgae
    ├── checkpoint.tar
    ├── embedding.npy
    ├── specter_embedding.npy
    ├── train_pos_edge_list.csv
    ├── test_pos_edge_list.csv
    └── test_neg_edge_list.csv

where the checkpoint.tar is the pretrained models' checkpoints. The specter_embedding.npy is the paper embeddings extracted by the Specter model, and the embedding.npy is the embeddings extracted by our VGAE model. train_pos_edge_list.csv, test_pos_edge_list.csv and test_neg_edge_list.csv are the training and testing edges of the VGAE model.

Note. If you want to train your own model from begining, please keep stats folder clean, otherwise the training scripts will automatically load the checkpoints in the stats and continue training. If you want to fine-tune the model, you can put the checkpoints in the corresponding folder within the stats folder and then continue fine-tuning it.