luke-base_on5

This model is a fine-tuned version of studio-ousia/luke-base on the ontonotes5 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0732
• F1-type-match: 0.4835
• F1-partial: 0.4930
• F1-strict: 0.4695
• F1-exact: 0.4832

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	F1-type-match	F1-partial	F1-strict	F1-exact
0.0756	1.0	936	0.0644	0.3975	0.4054	0.3800	0.3941
0.0508	2.0	1873	0.0589	0.6196	0.6313	0.5967	0.6158
0.0347	3.0	2809	0.0664	0.4686	0.4772	0.4530	0.4665
0.0243	4.0	3746	0.0677	0.3951	0.4033	0.3830	0.3948
0.0166	5.0	4680	0.0732	0.4835	0.4930	0.4695	0.4832

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.14.1