config_pretrain.yaml

batch_size: 512 # batch size
warm_up: 10 # warm-up epochs
epochs: 2 # total number of epochs

load_model: pretrained_gin # resume training
eval_every_n_epochs: 1 # validation frequency
save_every_n_epochs: 5 # automatic model saving frequecy
log_every_n_steps: 50 # print training log frequency

fp16_precision: False # float precision 16 (i.e. True/False)
init_lr: 0.0005 # initial learning rate for Adam
weight_decay: 1e-5 # weight decay for Adam
gpu: cuda:0 # training GPU

model:
  num_layer: 5 # number of graph conv layers
  emb_dim: 300 # embedding dimension in graph conv layers
  feat_dim: 512 # output feature dimention
  drop_ratio: 0 # dropout ratio
  pool: mean # readout pooling (i.e., mean/max/add)

aug: node # molecule graph augmentation strategy (i.e., node/subgraph/mix)
dataset:
  num_workers: 12 # dataloader number of workers
  valid_size: 0.05 # ratio of validation data
  data_path: data/{yourowndata}.csv # path of pre-training data

loss:
  temperature: 0.1 # temperature of NT-Xent loss
  use_cosine_similarity: True # whether to use cosine similarity in NT-Xent loss (i.e. True/False)