Downloads: 1
Diff of
/medicalbert/cliparser.py
[000000]
..
[d129b2]
Switch to unified view
| a | b/medicalbert/cliparser.py | ||
|---|---|---|---|
| 1 | import argparse |
||
| 2 | |||
| 3 | #All the parameters that we can set. |
||
| 4 | # NB: not all params are used by every classifier. |
||
| 5 | def setup_parser(): |
||
| 6 | parser = argparse.ArgumentParser() |
||
| 7 | |||
| 8 | parser.add_argument("--train_from_checkpoint", |
||
| 9 | default=None, |
||
| 10 | type=str, |
||
| 11 | help="Continue training from a saved model.") |
||
| 12 | |||
| 13 | parser.add_argument("--save_tokenized_text", |
||
| 14 | action='store_true', |
||
| 15 | help="this will output the tokenized process text into a CSV format") |
||
| 16 | parser.add_argument("--train", |
||
| 17 | action='store_true', |
||
| 18 | help="Whether to run training.") |
||
| 19 | parser.add_argument("--output_embeddings", |
||
| 20 | action='store_true', |
||
| 21 | help="Will take in a classifier and use the underlying model to output the token embeddings") |
||
| 22 | parser.add_argument("--eval", |
||
| 23 | action='store_true', |
||
| 24 | help="Whether to run eval on the dev set.") |
||
| 25 | parser.add_argument("--use_model", |
||
| 26 | default=None, |
||
| 27 | type=str, |
||
| 28 | help="Use this model for evaluations") |
||
| 29 | parser.add_argument("--data_dir", |
||
| 30 | default=None, |
||
| 31 | type=str, |
||
| 32 | help="location of input data") |
||
| 33 | parser.add_argument("--output_dir", |
||
| 34 | default=None, |
||
| 35 | type=str, |
||
| 36 | help="location of output") |
||
| 37 | parser.add_argument("--training_data", |
||
| 38 | default=None, |
||
| 39 | type=str, |
||
| 40 | help="name of training file") |
||
| 41 | parser.add_argument("--validation_metric", |
||
| 42 | default=None, |
||
| 43 | type=str, |
||
| 44 | help="metric used to select the best validation checkpoint for testing.") |
||
| 45 | parser.add_argument("--valid_data", |
||
| 46 | default=None, |
||
| 47 | type=str, |
||
| 48 | help="name of validation file") |
||
| 49 | parser.add_argument("--evaluator", |
||
| 50 | default=None, |
||
| 51 | type=str, |
||
| 52 | help="evaluation class to use") |
||
| 53 | parser.add_argument("--seed", |
||
| 54 | default=None, |
||
| 55 | type=int, |
||
| 56 | help="random seed") |
||
| 57 | parser.add_argument("--device", |
||
| 58 | default=None, |
||
| 59 | type=str, |
||
| 60 | help="cpu or cuda") |
||
| 61 | parser.add_argument("--experiment_name", |
||
| 62 | default=None, |
||
| 63 | type=str, |
||
| 64 | help="name of the experiment") |
||
| 65 | parser.add_argument("--learning_rate", |
||
| 66 | default=None, |
||
| 67 | type=float, |
||
| 68 | help="learning_rate") |
||
| 69 | parser.add_argument("--pretrained_model", |
||
| 70 | default=None, |
||
| 71 | type=str, |
||
| 72 | help="pretrained model to train upon.") |
||
| 73 | parser.add_argument("--num_sections", |
||
| 74 | default=None, |
||
| 75 | type=int, |
||
| 76 | help="chunks of text") |
||
| 77 | parser.add_argument("--tokenizer", |
||
| 78 | default=None, |
||
| 79 | type=str, |
||
| 80 | help="tokenizer model to use") |
||
| 81 | parser.add_argument("--num_train_examples", |
||
| 82 | default=None, |
||
| 83 | type=int, |
||
| 84 | help="number of training examples") |
||
| 85 | parser.add_argument("--target", |
||
| 86 | default=None, |
||
| 87 | type=str, |
||
| 88 | help="target column") |
||
| 89 | parser.add_argument("--classifier", |
||
| 90 | default=None, |
||
| 91 | type=str, |
||
| 92 | help="classifier to use") |
||
| 93 | parser.add_argument("--epochs", |
||
| 94 | default=None, |
||
| 95 | type=int, |
||
| 96 | help="Number of epochs to train for") |
||
| 97 | parser.add_argument("--train_batch_size", |
||
| 98 | default=None, |
||
| 99 | type=int, |
||
| 100 | help="batch size during training phase") |
||
| 101 | parser.add_argument("--gradient_accumulation_steps", |
||
| 102 | default=None, |
||
| 103 | type=int, |
||
| 104 | help="used to reduce GPU memory footprint") |
||
| 105 | parser.add_argument("--datareader", |
||
| 106 | default=None, |
||
| 107 | type=str, |
||
| 108 | help="approach to reading the data from files.") |
||
| 109 | parser.add_argument("--vocab_size", |
||
| 110 | default=None, |
||
| 111 | type=int, |
||
| 112 | help="Size of vocabulary.") |
||
| 113 | parser.add_argument("--embed_size", |
||
| 114 | default=None, |
||
| 115 | type=int, |
||
| 116 | help="Size of vocabulary.") |
||
| 117 | parser.add_argument("--layer", |
||
| 118 | default=None, |
||
| 119 | type=int, |
||
| 120 | help="If the classifier only uses parts of a model then use this") |
||
| 121 | parser.add_argument("--max_sequence_length", |
||
| 122 | default=None, |
||
| 123 | type=int, |
||
| 124 | help="maximum sequence length, each document will be truncated to this length.") |
||
| 125 | parser.add_argument("--num_layers", |
||
| 126 | default=None, |
||
| 127 | type=int, |
||
| 128 | help="The number of encoding layers for a BERT model to keep.") |
||
| 129 | return parser.parse_args() |