{
  "acc_batch_size": -1,
  "adam_betas": "0.9,0.999",
  "architectures": [
    "TransformerForNLI"
  ],
  "attention_probs_dropout_prob": 0.1,
  "bias_proj": "",
  "bias_v1": "",
  "bias_v2": "",
  "bos_token_id": 0,
  "clip": 1.0,
  "cls": "linear",
  "dir": "data/snli_1.0/",
  "dropout": 0.1,
  "enc": "transformer",
  "eos_token_id": 2,
  "epochs": 5,
  "freeze_transformer": 0,
  "gpuid": 1,
  "gradient_checkpointing": false,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 768,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label_dict": "data/snli_1.0/snli.label.dict",
  "label_map_inv": {
    "0": "entailment",
    "1": "neutral",
    "2": "contradiction"
  },
  "labels": [
    "entailment",
    "neutral",
    "contradiction"
  ],
  "layer_norm_eps": 1e-05,
  "learning_rate": 3e-05,
  "load_file": "",
  "loss": "multiclass",
  "max_position_embeddings": 514,
  "model_type": "transformerfornli",
  "num_attention_heads": 12,
  "num_hidden_layers": 12,
  "num_label": 3,
  "optim": "adamw_fp16",
  "pad_token_id": 1,
  "param_init_type": "xavier_normal",
  "percent": 1.0,
  "print_every": 500,
  "save_file": "./models/robertabase_snli_seed1",
  "seed": 1,
  "train_data": "data/snli_1.0/snli.train.hdf5",
  "train_res": "",
  "transformer_type": "roberta-base",
  "type_vocab_size": 1,
  "val_data": "data/snli_1.0/snli.val.hdf5",
  "val_res": "",
  "vocab_size": 50265,
  "warmup": "linear",
  "warmup_perc": 0.1,
  "weight_decay": 0.01
}