jialicheng/ddi-pubmedbert-fulltext

pubmedbert-fulltext

This model is a fine-tuned version of microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4548
• Accuracy: 0.9498

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 256
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	791	0.2502	0.9342
0.1717	2.0	1582	0.2889	0.9449
0.0792	3.0	2373	0.2844	0.9424
0.0565	4.0	3164	0.3055	0.9377
0.0565	5.0	3955	0.3059	0.9458
0.0405	6.0	4746	0.3693	0.9451
0.0274	7.0	5537	0.3295	0.9438
0.0263	8.0	6328	0.4278	0.9337
0.0181	9.0	7119	0.3807	0.9465
0.0181	10.0	7910	0.4318	0.9442
0.0173	11.0	8701	0.3995	0.9487
0.011	12.0	9492	0.4487	0.9466
0.0077	13.0	10283	0.4247	0.9482
0.0075	14.0	11074	0.5082	0.9433
0.0075	15.0	11865	0.4722	0.9458
0.0071	16.0	12656	0.4134	0.9507
0.0034	17.0	13447	0.4252	0.9496
0.0033	18.0	14238	0.4436	0.9500
0.0023	19.0	15029	0.4481	0.9505
0.0023	20.0	15820	0.4548	0.9498

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu118
• Datasets 2.18.0
• Tokenizers 0.15.2