ner_ontonotes.json

{
  "dataset_reader": {
    "name": "ontonotes_reader",
    "data_path": "ontonotes_senna"
  },
  "dataset_iterator": {
    "name": "data_learning_iterator",
    "seed": 42
  },
  "chainer": {
    "in": ["x"],
    "in_y": ["y"],
    "pipe": [
      {
        "in": ["x"],
        "name": "lazy_tokenizer",
        "out": ["x_tokens"]
      },
      {
        "in": ["x_tokens"],
        "name": "str_lower",
        "out": ["x_lower"]
      },
      {
        "in": ["x_lower"],
        "name": "sanitizer",
        "nums": true,
        "out": ["x_san"]
      },
      {
        "in": ["y"],
        "id": "tag_vocab",
        "name": "simple_vocab",
        "pad_with_zeros": true,
        "fit_on": ["y"],
        "save_path": "ner_ontonotes/tag.dict",
        "load_path": "ner_ontonotes/tag.dict",
        "out": ["y_ind"]
      },
      {
        "in": ["x_tokens"],
        "name": "char_splitter",
        "out": ["x_char"]
      },
      {
        "in": ["x_char"],
        "id": "char_vocab",
        "name": "char_vocab",
        "pad_with_zeros": true,
        "fit_on": ["x_char"],
        "save_path": "ner_ontonotes/char.dict",
        "load_path": "ner_ontonotes/char.dict",
        "out": ["x_char_ind"]
      },
      {
        "in": ["x_tokens"],
        "name": "mask",
        "out": ["mask"]
      },
      {
        "in": ["x_san"],
        "id": "glove_emb",
        "name": "glove",
        "pad_zero": true,
        "load_path": "embeddings/glove.6B.100d.txt",

        "out": ["x_emb"]
      },
      {
        "id": "embeddings_char",
        "name": "emb_mat_assembler",
        "character_level": true,
        "emb_dim": 32,
        "embedder": "#glove_emb",
        "vocab": "#char_vocab"
      },
      {
        "id": "capitalization",
        "name": "capitalization_featurizer",
        "in": ["x_tokens"],
        "out": ["cap"]
      },
      {
        "in": ["x_emb", "mask", "x_char_ind", "cap"],
        "in_y": ["y_ind"],
        "out": ["y_predicted"],
        "name": "ner",
        "main": true,
        "token_emb_dim": "#glove_emb.dim",
        "n_hidden_list": [256, 256, 256],
        "net_type": "rnn",
        "cell_type": "lstm",
        "use_cudnn_rnn": true,
        "n_tags": "#tag_vocab.len",
        "capitalization_dim": "#capitalization.dim",
        "char_emb_dim": "#embeddings_char.dim",
        "save_path": "ner_ontonotes/model",
        "load_path": "ner_ontonotes/model",
        "char_emb_mat": "#embeddings_char.emb_mat",
        "two_dense_on_top": true,
        "use_crf": true,
        "use_batch_norm": true,
        "embeddings_dropout": true,
        "top_dropout": true,
        "intra_layer_dropout": false,
        "l2_reg": 0,
        "learning_rate": 3e-3,
        "dropout_keep_prob": 0.7
      },
      {
        "ref": "tag_vocab",
        "in": ["y_predicted"],
        "out": ["tags"]
      }
    ],

    "out": ["x_tokens", "tags"]
  },
  "train": {
    "epochs": 100,
    "batch_size": 64,

    "metrics": ["ner_f1"],
    "validation_patience": 7,
    "val_every_n_epochs": 1,

    "log_every_n_epochs": -1,
    "show_examples": false
  },
  "metadata": {
    "requirements": [
      "../dp_requirements/gensim.txt",
      "../dp_requirements/tf-gpu.txt"
    ],
    "labels": {
      "telegram_utils": "NERCoNLL2003Model",
      "server_utils": "NER"
    },
    "download": [
      "http://files.deeppavlov.ai/deeppavlov_data/ner_ontonotes_v3_cpu_compatible.tar.gz",
      {
        "url": "http://files.deeppavlov.ai/embeddings/glove.6B.100d.txt",
        "subdir": "embeddings"
      }
    ]
  }

}