nb-bert-large-user-needs-v2

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

• Loss: 1.0173
• Accuracy: 0.8
• F1: 0.7945
• Precision: 0.7947
• Recall: 0.8

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: 8
• 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	188	0.7673	0.696	0.6619	0.6566	0.696
No log	2.0	376	0.5713	0.7707	0.7423	0.7163	0.7707
0.6847	3.0	564	0.5849	0.7653	0.7547	0.7654	0.7653
0.6847	4.0	752	0.7731	0.7467	0.7254	0.7474	0.7467
0.6847	5.0	940	0.6056	0.7733	0.7740	0.7756	0.7733
0.4443	6.0	1128	0.7752	0.792	0.7877	0.7901	0.792
0.4443	7.0	1316	1.0173	0.8	0.7945	0.7947	0.8
0.2807	8.0	1504	1.1683	0.7813	0.7789	0.7783	0.7813
0.2807	9.0	1692	1.1886	0.7893	0.7825	0.7841	0.7893
0.2807	10.0	1880	1.3052	0.776	0.7695	0.7729	0.776
0.1282	11.0	2068	1.4641	0.784	0.7769	0.7804	0.784
0.1282	12.0	2256	1.5614	0.7813	0.7716	0.7871	0.7813
0.1282	13.0	2444	1.6424	0.784	0.7774	0.7804	0.784
0.0529	14.0	2632	1.7250	0.7813	0.7767	0.7770	0.7813
0.0529	15.0	2820	1.6061	0.8	0.7934	0.8058	0.8
0.0182	16.0	3008	1.7678	0.792	0.7854	0.7908	0.792
0.0182	17.0	3196	1.8226	0.7893	0.7834	0.7849	0.7893
0.0182	18.0	3384	1.8330	0.7973	0.7906	0.7936	0.7973
0.0061	19.0	3572	1.8423	0.7947	0.7879	0.7909	0.7947
0.0061	20.0	3760	1.8536	0.7973	0.7906	0.7936	0.7973

Framework versions

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