scibert-lm-const-finetuned-20

This model is a fine-tuned version of allenai/scibert_scivocab_cased on the conll2003 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0099

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: 4
• eval_batch_size: 4
• 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
3.6081	1.0	118	2.9156
2.7954	2.0	236	2.5940
2.5762	3.0	354	2.5017
2.4384	4.0	472	2.3923
2.3391	5.0	590	2.2996
2.2417	6.0	708	2.3180
2.2161	7.0	826	2.2336
2.1918	8.0	944	2.2465
2.1494	9.0	1062	2.1871
2.1215	10.0	1180	2.1566
2.1015	11.0	1298	2.1849
2.05	12.0	1416	2.1092
2.0653	13.0	1534	2.2221
2.0261	14.0	1652	2.1572
2.0117	15.0	1770	2.1452
1.9845	16.0	1888	2.1433
1.9791	17.0	2006	2.1225
1.9979	18.0	2124	2.0777
1.9688	19.0	2242	2.1765
1.9873	20.0	2360	2.0099

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0+cu113
• Datasets 2.3.2
• Tokenizers 0.12.1