jialicheng/ddi-pubmedbert-abstract

pubmedbert-abstract

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

• Loss: 0.4618
• Accuracy: 0.9501

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	0.2340	0.9386
0.1776	2.0	1582	0.2716	0.9419
0.0855	3.0	2373	0.2730	0.9431
0.0627	4.0	3164	0.3323	0.9382
0.0627	5.0	3955	0.3308	0.9451
0.0463	6.0	4746	0.3986	0.9412
0.0308	7.0	5537	0.4211	0.9419
0.0312	8.0	6328	0.3616	0.9437
0.0221	9.0	7119	0.4310	0.9396
0.0221	10.0	7910	0.4222	0.9438
0.0181	11.0	8701	0.4185	0.9445
0.0141	12.0	9492	0.4678	0.9456
0.0133	13.0	10283	0.4027	0.9503
0.0082	14.0	11074	0.4504	0.9473
0.0082	15.0	11865	0.4760	0.9505
0.0052	16.0	12656	0.4573	0.9449
0.0042	17.0	13447	0.4356	0.9522
0.0037	18.0	14238	0.4577	0.9487
0.0024	19.0	15029	0.4642	0.9493
0.0024	20.0	15820	0.4618	0.9501

Framework versions

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