norbert3-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: 3.1392
• Accuracy: 0.7067
• F1: 0.6946
• Precision: 0.6905
• Recall: 0.7067

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	0.8059	0.6747	0.6472	0.6569	0.6747
0.9129	2.0	750	0.9030	0.6453	0.6142	0.5975	0.6453
0.7636	3.0	1125	0.7755	0.6667	0.6292	0.6250	0.6667
0.6003	4.0	1500	1.0267	0.6773	0.6591	0.6928	0.6773
0.6003	5.0	1875	1.9897	0.6267	0.6378	0.6526	0.6267
0.2905	6.0	2250	2.0507	0.704	0.6913	0.6879	0.704
0.0901	7.0	2625	2.7638	0.6853	0.6590	0.6863	0.6853
0.0365	8.0	3000	2.6138	0.696	0.6875	0.6907	0.696
0.0365	9.0	3375	3.0024	0.6667	0.6585	0.6543	0.6667
0.0162	10.0	3750	2.9416	0.6933	0.6829	0.6798	0.6933
0.0022	11.0	4125	3.2015	0.6827	0.6558	0.6790	0.6827
0.0114	12.0	4500	3.3133	0.6933	0.6694	0.6916	0.6933
0.0114	13.0	4875	3.2376	0.6773	0.6695	0.6647	0.6773
0.0042	14.0	5250	3.1392	0.7067	0.6946	0.6905	0.7067
0.0035	15.0	5625	3.2710	0.6907	0.6770	0.6705	0.6907
0.0045	16.0	6000	3.3476	0.6933	0.6847	0.6841	0.6933
0.0045	17.0	6375	3.2386	0.696	0.6904	0.6932	0.696
0.0065	18.0	6750	3.4263	0.6853	0.6700	0.6607	0.6853
0.0029	19.0	7125	3.4898	0.6827	0.6652	0.6579	0.6827
0.0013	20.0	7500	3.5103	0.68	0.6624	0.6554	0.68

Framework versions

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