training_config.json

{
  "seed": 42,
  "dataset_name": "pancreatic_full",
  "do_train": true,
  "load_model": null,
  "GEPC": true,
  "ecs_thres": 0.7,
  "dab_weight": 0,
  "this_weight": 1,
  "next_weight": 0,
  "n_rounds": 3,
  "next_cell_pred_type": "identity",
  "ecs_weight": 1,
  "cell_type_classifier": true,
  "cell_type_classifier_weight": 1,
  "perturbation_classifier_weight": 5,
  "perturbation_input": false,
  "CCE": true,
  "mask_ratio": 0.15,
  "epochs": 150,
  "lr": 0.0003,
  "batch_size": 128,
  "layer_size": 256,
  "schedule_ratio": 0.995,
  "save_eval_interval": 5,
  "log_interval": 200,
  "fast_transformer": true,
  "amp": true,
  "do_sample_in_train": false,
  "ADV": false,
  "adv_weight": 10000,
  "adv_E_delay_epochs": 2,
  "adv_D_delay_epochs": 2,
  "lr_ADV": 0.001,
  "DSBN": false,
  "per_seq_batch_sample": false,
  "use_batch_label": false,
  "schedule_interval": 1,
  "cce_mode": "cell+geno",
  "n_hvg": 2000,
  "ps_weight": 0,
  "mask_value": -1,
  "cls_value": -3,
  "cls_token": "<cls>",
  "sampling_mode": "hvg",
  "non_hvg_size": 1000,
  "fix_nonzero_prop": true,
  "nonzero_prop": 0.9,
  "hvg_col": "highly_variable",
  "mvc_masked_train": true,
  "max_seq_len": 3001,
  "n_pert": 40,
  "nlayers_pert": 4,
  "n_ps": 1,
  "ntoken": 36604,
  "d_model": 256,
  "nhead": 8,
  "d_hid": 256,
  "nlayers": 6,
  "nlayers_cls": 3,
  "n_cls": 20,
  "vocab": null,
  "dropout": 0.1,
  "pad_token": "<pad>",
  "pad_value": -2,
  "do_mvc": true,
  "do_dab": false,
  "use_batch_labels": false,
  "num_batch_labels": null,
  "domain_spec_batchnorm": false,
  "input_emb_style": "continuous",
  "n_bins": 51,
  "cell_emb_style": "cls",
  "mvc_decoder_style": "inner product",
  "ecs_threshold": 0.7,
  "explicit_zero_prob": true,
  "use_fast_transformer": true,
  "fast_transformer_backend": "flash",
  "pre_norm": false,
  "pred_lochness_next": false,
  "ps_decoder2_nlayer": 3,
  "pert_pad_id": null,
  "pert_dim": null
}