pilotj/newsense-roberta-base

roberta-base-v1

This model is a fine-tuned version of pilotj/roberta-base-pretrained-v1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3920
• Accuracy: 0.8867
• F1 Macro: 0.8576
• F1 W: 0.8880
• Precision: 0.8909
• Recall: 0.8867

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: 128
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 W	Precision	Recall
0.3932	0.1896	500	0.4138	0.8803	0.8505	0.8816	0.8847	0.8803
0.3997	0.3792	1000	0.4097	0.8809	0.8499	0.8824	0.8861	0.8809
0.3997	0.5688	1500	0.4126	0.8818	0.8514	0.8834	0.8874	0.8818
0.3907	0.7584	2000	0.3988	0.8844	0.8544	0.8856	0.8887	0.8844
0.3881	0.9480	2500	0.3956	0.8862	0.8549	0.8871	0.8901	0.8862
0.3558	1.1377	3000	0.3971	0.8863	0.8570	0.8874	0.8902	0.8863
0.3526	1.3273	3500	0.3999	0.8852	0.8558	0.8867	0.8902	0.8852
0.3435	1.5169	4000	0.3991	0.8858	0.8565	0.8870	0.8903	0.8858
0.3428	1.7065	4500	0.3929	0.8859	0.8572	0.8871	0.8901	0.8859
0.3392	1.8961	5000	0.3920	0.8867	0.8576	0.8880	0.8909	0.8867

Framework versions

• Transformers 4.45.1
• Pytorch 2.2.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0