nb-bert-large-user-needs-v2

This model is a fine-tuned version of ltg/norbert3-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.0724
• Accuracy: 0.6853
• F1: 0.6674
• Precision: 0.6636
• Recall: 0.6853

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: 3e-05
• train_batch_size: 4
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
No log	1.0	375	0.8229	0.6453	0.6076	0.6107	0.6453
0.9043	2.0	750	0.8588	0.6667	0.6416	0.6207	0.6667
0.744	3.0	1125	0.8531	0.656	0.6300	0.6063	0.656
0.5288	4.0	1500	1.5591	0.6827	0.6513	0.6425	0.6827
0.5288	5.0	1875	2.0799	0.6507	0.6495	0.6493	0.6507
0.1999	6.0	2250	2.7613	0.664	0.6465	0.6476	0.664
0.0406	7.0	2625	3.0547	0.6267	0.6044	0.6314	0.6267
0.0402	8.0	3000	2.7554	0.672	0.6613	0.6515	0.672
0.0402	9.0	3375	2.9426	0.6587	0.6529	0.6478	0.6587
0.0229	10.0	3750	3.0100	0.664	0.6505	0.6462	0.664
0.0075	11.0	4125	2.9582	0.656	0.6375	0.6266	0.656
0.0112	12.0	4500	2.9971	0.656	0.6424	0.6335	0.656
0.0112	13.0	4875	3.1416	0.6587	0.6422	0.6297	0.6587
0.0054	14.0	5250	3.1233	0.68	0.6597	0.6495	0.68
0.003	15.0	5625	3.1611	0.6773	0.6632	0.6657	0.6773
0.0071	16.0	6000	3.0724	0.6853	0.6674	0.6636	0.6853
0.0071	17.0	6375	3.0737	0.6667	0.6538	0.6468	0.6667
0.0029	18.0	6750	3.1641	0.6747	0.6659	0.6599	0.6747
0.0024	19.0	7125	3.2062	0.6693	0.6597	0.6542	0.6693
0.0012	20.0	7500	3.2474	0.6747	0.6650	0.6595	0.6747

Framework versions

• Transformers 4.36.0
• Pytorch 2.1.0
• Datasets 2.18.0
• Tokenizers 0.15.2