config.json

{
  "pipeline": {
    "preprocess": true,
    "train": true,
    "explain": true,
    "postprocess": true
  },
  "preprocessing": {
    "word_vectors": {
      "doit": true,
      "file_word2vec": "./data/embeddings/wiki-news-300d-1M.vec",
      "dir_vocab": "./data/embeddings/",
      "vocab_size": 50000
    },
    "pubmed": {
      "doit": true,
      "to_lower": false,
      "file_train_text": "./data/PubMed_20k_RCT/train.txt",
      "file_dev_text": "./data/PubMed_20k_RCT/dev.txt",
      "file_test_text": "./data/PubMed_20k_RCT/test.txt",
      "language_model": "en_core_sci_sm"
    }
  },
  "training": {
    "doit": true,
    "pad": 350,
    "crop_train": -1,
    "crop_dev": -1,
    "crop_test": -1,
    "batch_size": 128,
    "num_workers": 8,
    "path_model": "./data/model/model.weights",
    "epochs": 100,
    "nclasses": 5,
    "nfeat": 300,
    "nhid": 300,
    "patience": 3,
    "metric": "weighted",
    "random_seed": 42
  },
  "explain": {
    "file_explanations_jsonl": "./data/explanations/explanations.jsonl",
    "crop": -1,
    "occlude": true,
    "drop": 1.0,
    "step": 0.1,
    "verbose": false
  },
  "postprocess": {
    "occlusion_experiment": {
      "doit": true,
      "path_out_top_masked_predictions": "./data/explanations/top_masked_predictions.csv",
      "path_out_bottom_masked_predictions": "./data/explanations/bottom_masked_predictions.csv",
      "draw_plot": false
    },
    "latex": {
      "doit": true,
      "path_out_latex": "./data/explanations/explanations.tex",
      "max_seq_len": 10,
      "crop": 250,
      "weight": 15,
      "base": 0.5
    }
  }
}