legal_bert_sm_cv_defined_4

This model is a fine-tuned version of nlpaueb/legal-bert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5637
• Accuracy: 0.817
• Precision: 0.5545
• Recall: 0.3128
• F1: 0.4
• D-index: 1.5629

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 8000
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	D-index
No log	1.0	250	0.4802	0.805	0.0	0.0	0.0	1.4370
0.5519	2.0	500	0.4506	0.809	0.8333	0.0256	0.0498	1.4518
0.5519	3.0	750	0.4347	0.828	0.7018	0.2051	0.3175	1.5411
0.4107	4.0	1000	0.4288	0.838	0.7037	0.2923	0.4130	1.5841
0.4107	5.0	1250	0.4980	0.834	0.7736	0.2103	0.3306	1.5510
0.3108	6.0	1500	0.4671	0.837	0.7353	0.2564	0.3802	1.5707
0.3108	7.0	1750	0.4817	0.835	0.6829	0.2872	0.4043	1.5784
0.2129	8.0	2000	0.6713	0.826	0.6981	0.1897	0.2984	1.5331
0.2129	9.0	2250	0.7290	0.825	0.6190	0.2667	0.3728	1.5581
0.1289	10.0	2500	0.8754	0.812	0.5402	0.2410	0.3333	1.5317
0.1289	11.0	2750	1.1026	0.822	0.6232	0.2205	0.3258	1.5383
0.0804	12.0	3000	1.2274	0.807	0.5109	0.2410	0.3275	1.5249
0.0804	13.0	3250	1.2411	0.824	0.5772	0.3641	0.4465	1.5894
0.0529	14.0	3500	1.2761	0.812	0.5263	0.3590	0.4268	1.5716
0.0529	15.0	3750	1.3524	0.823	0.5714	0.3692	0.4486	1.5897
0.0299	16.0	4000	1.6109	0.829	0.7308	0.1949	0.3077	1.5389
0.0299	17.0	4250	1.6461	0.824	0.6863	0.1795	0.2846	1.5269
0.0284	18.0	4500	1.7304	0.824	0.7879	0.1333	0.2281	1.5108
0.0284	19.0	4750	1.5481	0.809	0.5159	0.3333	0.4050	1.5590
0.0396	20.0	5000	1.5637	0.817	0.5545	0.3128	0.4	1.5629

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.1+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3