nci-ner-v2-staging

This model is a fine-tuned version of FacebookAI/roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0727
• Precision: 0.8815
• Recall: 0.9097
• F1: 0.8954

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1
0.2779	0.5339	500	0.1139	0.8024	0.8149	0.8086
0.1683	1.0673	1000	0.0903	0.8271	0.8766	0.8512
0.1431	1.6012	1500	0.0836	0.8484	0.8855	0.8666
0.1106	2.1345	2000	0.0833	0.8535	0.8895	0.8711
0.0950	2.6684	2500	0.0879	0.8532	0.8895	0.8710
0.0832	3.2018	3000	0.0832	0.8624	0.8919	0.8769
0.0728	3.7357	3500	0.0833	0.8619	0.8925	0.8769
0.0688	4.2691	4000	0.0853	0.8599	0.8956	0.8774
0.0734	4.8030	4500	0.0846	0.8610	0.8956	0.8780
0.0667	5.0	4685	0.0847	0.8607	0.8953	0.8777

Framework versions

• Transformers 5.15.1
• Pytorch 2.13.0+cu130
• Datasets 4.8.5
• Tokenizers 0.22.2