wongnai.json

{
    "dataset_reader": {
      "type": "wongnai_reader", 
      "lazy": true
    },
    "train_data_path": "wongnai_dataset/training.jsonl",
    "validation_data_path": "wongnai_dataset/validation.jsonl",
    "model": {
      "type": "wongnai_classifier",
      "text_field_embedder": {
        "tokens": {
          "type": "embedding",
          "pretrained_file": "https://s3-us-west-2.amazonaws.com/thai-corpus/thai2vec02.txt.gz",
          "embedding_dim": 300,
          "trainable": false
        }
      },
      "comment_encoder": {
        "type": "lstm",
        "bidirectional": true,
        "input_size": 300,
        "hidden_size": 300,
        "num_layers": 1,
        "dropout": 0.2
      },
      "classifier_feedforward": {
        "input_dim": 600,
        "num_layers": 2,
        "hidden_dims": [600, 5],
        "activations": ["relu", "linear"],
        "dropout": [0.2, 0.0]
      }
    },
    "iterator": {
      "type": "bucket",
      "sorting_keys": [["comment", "num_tokens"]],
      "batch_size": 128
    },
    "trainer": {
      "num_epochs": 50,
      "patience": 10,
      "cuda_device": 0,
      "grad_clipping": 5.0,
      "validation_metric": "+accuracy",
      "optimizer": {
        "type": "adagrad"
      }
    }
}