snips.json

{
  "task": "tagger",
  "batchsz": 10,
  "conll_output": "snips.conll",
  "basedir": "./snips-sf",
  "unif": 0.1,
  "features": [
    {
      "name": "word",
      "vectorizer": {
        "type": "token1d",
        "transform": "baseline.lowercase"
      },
      "embeddings": {
        "label": "glove-6B-100"
      }
    },
    {
      "name": "senna",
      "vectorizer": {
        "type": "token1d",
        "transform": "baseline.lowercase"
      },
      "embeddings": {
        "label": "senna"
      }
    },
    {
      "name": "char",
      "vectorizer": {
        "type": "char2d"
      },
      "embeddings": { "dsz": 30, "wsz": 30, "type": "char-conv" }
    }
  ],
  "preproc": {
  },
  "backend": "pytorch",
  "dataset": "snips",
  "loader": {
    "reader_type": "parallel",
    "data_suffix": "in",
    "tag_suffix": "out"
  },
  "model": {
    "model_type": "default",
    "cfiltsz": [
      3
    ],
    "hsz": 400,
    "dropout": 0.5,
    "dropin": {"word": 0.1,"senna": 0.1},
    "rnntype": "blstm",
    "layers": 1,
    "constrain_decode": true,
    "crf": 1
  },
  "train": {
    "epochs": 100,
    "optim": "sgd",
    "eta": 0.015,
    "mom": 0.9,
    "patience": 40,
    "early_stopping_metric": "f1",
    "clip": 5.0,
    "span_type": "iobes"
  }
}