README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: BERT-evidence-types
    results: []

BERT-evidence-types

This model is a fine-tuned version of bert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.4683
• Macro f1: 0.3854
• Weighted f1: 0.6985
• Accuracy: 0.7116
• Balanced accuracy: 0.3720

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Macro f1	Weighted f1	Accuracy	Balanced accuracy
1.1736	1.0	125	1.0715	0.2535	0.6525	0.6667	0.2721
0.8182	2.0	250	0.9876	0.3068	0.6769	0.6804	0.3152
0.6024	3.0	375	1.0436	0.4067	0.7024	0.7047	0.4089
0.4004	4.0	500	1.1912	0.4045	0.6988	0.7078	0.4050
0.2531	5.0	625	1.3408	0.3882	0.6861	0.6903	0.3967
0.1625	6.0	750	1.5378	0.3866	0.6951	0.7040	0.3807
0.0985	7.0	875	1.7579	0.3850	0.6990	0.7161	0.3824
0.0664	8.0	1000	1.9837	0.3609	0.6849	0.7032	0.3529
0.0411	9.0	1125	2.0033	0.3807	0.6929	0.7024	0.3618
0.0262	10.0	1250	2.1714	0.3771	0.6924	0.7085	0.3585
0.0204	11.0	1375	2.2539	0.3734	0.6832	0.6933	0.3658
0.0141	12.0	1500	2.3033	0.3654	0.6830	0.6979	0.3556
0.0118	13.0	1625	2.3853	0.3679	0.6912	0.7108	0.3520
0.0109	14.0	1750	2.3749	0.3810	0.6952	0.7100	0.3665
0.0078	15.0	1875	2.4042	0.3777	0.6942	0.7078	0.3645
0.0079	16.0	2000	2.5097	0.3790	0.6938	0.7123	0.3632
0.0073	17.0	2125	2.4305	0.3844	0.6957	0.7070	0.3725
0.0046	18.0	2250	2.4700	0.3762	0.6941	0.7093	0.3638
0.0064	19.0	2375	2.4566	0.3844	0.6974	0.7100	0.3713
0.0057	20.0	2500	2.4683	0.3854	0.6985	0.7116	0.3720

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1