| { | |
| "dataset_reader": { | |
| "type": "drop", | |
| "instance_format": "drop", | |
| "passage_length_limit": 400, | |
| "question_length_limit": 50, | |
| "skip_when_all_empty": [ | |
| "passage_span", | |
| "question_span", | |
| "addition_subtraction", | |
| "counting" | |
| ], | |
| "token_indexers": { | |
| "token_characters": { | |
| "type": "characters", | |
| "min_padding_length": 5 | |
| }, | |
| "tokens": { | |
| "type": "single_id", | |
| "lowercase_tokens": true | |
| } | |
| } | |
| }, | |
| "model": { | |
| "type": "naqanet", | |
| "answering_abilities": [ | |
| "passage_span_extraction", | |
| "question_span_extraction", | |
| "addition_subtraction", | |
| "counting" | |
| ], | |
| "dropout_prob": 0.1, | |
| "matrix_attention_layer": { | |
| "type": "linear", | |
| "combination": "x,y,x*y", | |
| "tensor_1_dim": 128, | |
| "tensor_2_dim": 128 | |
| }, | |
| "modeling_layer": { | |
| "type": "qanet_encoder", | |
| "attention_dropout_prob": 0, | |
| "attention_projection_dim": 128, | |
| "conv_kernel_size": 5, | |
| "dropout_prob": 0.1, | |
| "feedforward_hidden_dim": 128, | |
| "hidden_dim": 128, | |
| "input_dim": 128, | |
| "layer_dropout_undecayed_prob": 0.1, | |
| "num_attention_heads": 8, | |
| "num_blocks": 6, | |
| "num_convs_per_block": 2 | |
| }, | |
| "num_highway_layers": 2, | |
| "phrase_layer": { | |
| "type": "qanet_encoder", | |
| "attention_dropout_prob": 0, | |
| "attention_projection_dim": 128, | |
| "conv_kernel_size": 7, | |
| "dropout_prob": 0.1, | |
| "feedforward_hidden_dim": 128, | |
| "hidden_dim": 128, | |
| "input_dim": 128, | |
| "layer_dropout_undecayed_prob": 0.1, | |
| "num_attention_heads": 8, | |
| "num_blocks": 1, | |
| "num_convs_per_block": 4 | |
| }, | |
| "regularizer": { | |
| "regexes": [ | |
| [ | |
| ".*", | |
| { | |
| "alpha": 1e-07, | |
| "type": "l2" | |
| } | |
| ] | |
| ] | |
| }, | |
| "text_field_embedder": { | |
| "token_embedders": { | |
| "token_characters": { | |
| "type": "character_encoding", | |
| "embedding": { | |
| "embedding_dim": 64 | |
| }, | |
| "encoder": { | |
| "type": "cnn", | |
| "embedding_dim": 64, | |
| "ngram_filter_sizes": [ | |
| 5 | |
| ], | |
| "num_filters": 200 | |
| } | |
| }, | |
| "tokens": { | |
| "type": "embedding", | |
| "embedding_dim": 300, | |
| "pretrained_file": "https://allennlp.s3.amazonaws.com/datasets/glove/glove.840B.300d.lower.converted.zip", | |
| "trainable": false | |
| } | |
| } | |
| } | |
| }, | |
| "train_data_path": "https://s3-us-west-2.amazonaws.com/allennlp/datasets/drop/drop_dataset.zip!drop_dataset/drop_dataset_train.json", | |
| "validation_data_path": "https://s3-us-west-2.amazonaws.com/allennlp/datasets/drop/drop_dataset.zip!drop_dataset/drop_dataset_dev.json", | |
| "trainer": { | |
| "callbacks": [ | |
| "tensorboard" | |
| ], | |
| "grad_norm": 5, | |
| "moving_average": { | |
| "type": "exponential", | |
| "decay": 0.9999 | |
| }, | |
| "num_epochs": 50, | |
| "optimizer": { | |
| "type": "adam", | |
| "betas": [ | |
| 0.8, | |
| 0.999 | |
| ], | |
| "eps": 1e-07, | |
| "lr": 0.0005 | |
| }, | |
| "patience": 10, | |
| "validation_metric": "+f1" | |
| }, | |
| "vocabulary": { | |
| "min_count": { | |
| "token_characters": 200 | |
| }, | |
| "only_include_pretrained_words": true, | |
| "pretrained_files": { | |
| "tokens": "https://allennlp.s3.amazonaws.com/datasets/glove/glove.840B.300d.lower.converted.zip" | |
| } | |
| }, | |
| "data_loader": { | |
| "batch_sampler": { | |
| "type": "bucket", | |
| "batch_size": 16 | |
| } | |
| }, | |
| "validation_dataset_reader": { | |
| "type": "drop", | |
| "instance_format": "drop", | |
| "passage_length_limit": 1000, | |
| "question_length_limit": 100, | |
| "skip_when_all_empty": [], | |
| "token_indexers": { | |
| "token_characters": { | |
| "type": "characters", | |
| "min_padding_length": 5 | |
| }, | |
| "tokens": { | |
| "type": "single_id", | |
| "lowercase_tokens": true | |
| } | |
| } | |
| } | |
| } |