train/models/transformer/big_config.json

{
  "emb_size": 1024,
  "feedforward_size": 4096,
  "hidden_size": 1024,
  "hidden_act": "relu",
  "heads_num": 16,
  "layers_num": 6,
  "decoder_layers_num": 6,
  "max_seq_length": 512,
  "dropout": 0.3,
  "data_processor": "mt",
  "embedding": ["word", "sinusoidalpos"],
  "tgt_embedding": ["word", "sinusoidalpos"],
  "encoder": "transformer",
  "decoder": "transformer",
  "mask": "fully_visible",
  "target": ["lm"],
  "tie_weights": true
}