nci-ner-v1

This model is a fine-tuned version of FacebookAI/roberta-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0693
• Precision: 0.8854
• Recall: 0.9097
• F1: 0.8974

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: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1
0.2002	0.2670	500	0.1542	0.6783	0.6954	0.6868
0.1159	0.5339	1000	0.0952	0.8101	0.8456	0.8275
0.0887	0.8009	1500	0.0805	0.8352	0.8625	0.8486
0.0653	1.0678	2000	0.0749	0.8421	0.8807	0.8610
0.0665	1.3348	2500	0.0734	0.8545	0.8754	0.8648
0.0558	1.6017	3000	0.0715	0.8565	0.8868	0.8714
0.0583	1.8687	3500	0.0689	0.8577	0.8899	0.8735
0.0453	2.1356	4000	0.0703	0.8689	0.8827	0.8758
0.0441	2.4026	4500	0.0713	0.8671	0.8894	0.8781
0.0409	2.6695	5000	0.0731	0.8562	0.8908	0.8731
0.0419	2.9365	5500	0.0702	0.8643	0.8935	0.8787
0.0379	3.2034	6000	0.0724	0.8602	0.8936	0.8766
0.0323	3.4704	6500	0.0745	0.8631	0.8931	0.8779
0.0303	3.7373	7000	0.0740	0.8659	0.8949	0.8802
0.0309	4.0043	7500	0.0726	0.8644	0.8991	0.8815
0.0258	4.2712	8000	0.0751	0.8650	0.8961	0.8803
0.0295	4.5382	8500	0.0753	0.8651	0.8963	0.8804
0.0292	4.8051	9000	0.0748	0.8694	0.8978	0.8833

Framework versions

• Transformers 4.46.3
• Pytorch 2.10.0+cu128
• Datasets 2.21.0
• Tokenizers 0.20.3