npt/config/mnist.json

{
  "model": {
    "dim_hidden": 32,
    "num_heads": 4,
    "stacking_depth": 4,
    "rff_depth": 1,
    "feature_type_embedding": false,
    "feature_index_embedding": true,
    "sep_res_embed": true,
    "mix_heads": true,
    "att_score_norm": "softmax",
    "embedding_layer_norm": true,
    "att_block_layer_norm": true,
    "pre_layer_norm": true,
    "layer_norm_eps": 1e-12,
    "hidden_dropout_prob": 0.1,
    "att_score_dropout_prob": 0.1,
    "bert_augmentation": true,
    "bert_mask_percentage": 0.9,
    "augmentation_bert_mask_prob": {
      "train": 0.15,
      "val": 0.3,
      "test": 0.15
    },
    "init_weights": true,
    "init_type": "xavier",
    "init_params": [
      "0.05",
      "0"
    ],
    "amp": true,
    "dtype": "float32",
    "is_semi_supervised": false
  },
  "training": {
    "batch_size": -1,
    "val_batchsize": 400,
    "num_train_inference": -1,
    "num_total_steps": 1000.0,
    "lr": 0.001,
    "weight_decay": 0.0,
    "gradient_clipping": 1.0,
    "optimizer": "lookahead_lamb",
    "scheduler": "flat_and_anneal",
    "lookahead_update_cadence": 6,
    "optimizer_warmup_proportion": 0.0,
    "optimizer_warmup_fixed_n_steps": 100,
    "minibatch_sgd": true,
    "eval_every_n": 1000,
    "eval_every_epoch_or_steps": "epochs",
    "eval_test_at_end_only": false,
    "mix_rows_inference": false,
    "checkpoint_setting": "all_checkpoints",
    "cache_cadence": 1,
    "checkpoint_save": 100000.0,
    "load_from_checkpoint": false,
    "patience": -1,
    "np_seed": 1,
    "torch_seed": 1,
    "original_np_seed": 91,
    "original_torch_seed": 1,
    "n_runs": 20
  },
  "data": {
    "data_path": "./data/mnist/",
    "name": "mnistad",
    "data_loader_nprocs": 0,
    "dataset_on_cuda": false,
    "data_force_reload": true,
    "data_log_mem_usage": false,
    "clear_tmp_files": false,
    "data_dtype": "float32",
    "val_perc": 0.1,
    "test_perc": 0.2,
    "keep_categorical_features": true,
    "cat_as_num_features": false,
    "num_reconstruction": null,
    "deterministic_masks": true,
    "n_hidden_features": [
      1
    ],
    "max_n_recon": 0,
    "aggregation": "sum",
    "normalize_ad_loss": false,
    "anomalies_in_inference": false,
    "full_trainset_inference": null,
    "contamination_share_train": 0,
    "share_contamination": 0.0
  },
  "system": {
    "use_cuda": true,
    "device": null,
    "distributed": true,
    "nodes": 1,
    "gpus": 4,
    "nr": -1,
    "no_sync": -1,
    "bucket_cap_mb": -1,
    "verbose": false,
    "print_every_nth_forward": false,
    "exp_name": "mnist_params",
    "checkpoint_key": "job__$Q_DATE"
  },
  "ad": true
}