hsohn3/ehr-bert-base-uncased-cchs-wordlevel

This model is a fine-tuned version of bert-base-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Train Loss: 3.7374
• Epoch: 9

Model description

• model: bert-base-uncased (train from scratch)
• tokenizer: BertTokenizer + WordLevel splitter

Intended uses & limitations

More information needed

Training and evaluation data

• data_source: cchs (10,000 visits)
• data_format: visit-level texts concatenated by [SEP] token

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'AdamWeightDecay', 'learning_rate': 1e-04, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False, 'weight_decay_rate': 0.01}
• training_precision: float32
• block_size: 512
• batch_size: 4
• num_epochs: 10
• mlm_probability: 0.15

Training results

Train Loss	Epoch
3.8857	0
3.7525	1
3.7505	2
3.7493	3
3.7412	4
3.7432	5
3.7428	6
3.7409	7
3.7394	8
3.7374	9

Framework versions

• Transformers 4.20.1
• TensorFlow 2.8.2
• Datasets 2.3.2
• Tokenizers 0.12.1