legal-bert-lora

This model is a fine-tuned version of nlpaueb/legal-bert-base-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6841
• Accuracy: 0.8048
• Precision: 0.7955
• Recall: 0.8048
• Precision Macro: 0.6332
• Recall Macro: 0.6316
• Macro Fpr: 0.0177
• Weighted Fpr: 0.0170
• Weighted Specificity: 0.9753
• Macro Specificity: 0.9853
• Weighted Sensitivity: 0.8048
• Macro Sensitivity: 0.6316
• F1 Micro: 0.8048
• F1 Macro: 0.6233
• F1 Weighted: 0.7978

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	Precision Macro	Recall Macro	Macro Fpr	Weighted Fpr	Weighted Specificity	Macro Specificity	Weighted Sensitivity	Macro Sensitivity	F1 Micro	F1 Macro	F1 Weighted
No log	1.0	160	1.2986	0.6421	0.5563	0.6421	0.2826	0.3627	0.0384	0.0383	0.9531	0.9730	0.6421	0.3627	0.6421	0.3114	0.5878
No log	2.0	321	0.8962	0.7273	0.6748	0.7273	0.3629	0.4471	0.0265	0.0261	0.9685	0.9797	0.7273	0.4471	0.7273	0.3889	0.6926
No log	3.0	482	0.7814	0.7413	0.7104	0.7413	0.3985	0.4561	0.0245	0.0243	0.9703	0.9808	0.7413	0.4561	0.7413	0.4041	0.7109
1.2548	4.0	643	0.7648	0.7382	0.7158	0.7382	0.4273	0.4496	0.0254	0.0247	0.9662	0.9803	0.7382	0.4496	0.7382	0.4122	0.7112
1.2548	5.0	803	0.7329	0.7452	0.7105	0.7452	0.4162	0.4569	0.0248	0.0238	0.9668	0.9808	0.7452	0.4569	0.7452	0.4120	0.7133
1.2548	6.0	964	0.7430	0.7568	0.7547	0.7568	0.4627	0.4868	0.0229	0.0224	0.9710	0.9819	0.7568	0.4868	0.7568	0.4504	0.7424
0.6432	7.0	1125	0.7300	0.7723	0.7524	0.7723	0.5180	0.5411	0.0213	0.0206	0.9724	0.9830	0.7723	0.5411	0.7723	0.5175	0.7578
0.6432	8.0	1286	0.7212	0.7699	0.7514	0.7699	0.5096	0.5397	0.0216	0.0209	0.9727	0.9828	0.7699	0.5397	0.7699	0.5123	0.7556
0.6432	9.0	1446	0.6910	0.7839	0.7634	0.7839	0.5217	0.5566	0.0200	0.0193	0.9728	0.9838	0.7839	0.5566	0.7839	0.5280	0.7690
0.4841	10.0	1607	0.7122	0.7878	0.7732	0.7878	0.5355	0.5777	0.0195	0.0189	0.9748	0.9842	0.7878	0.5777	0.7878	0.5495	0.7776
0.4841	11.0	1768	0.6813	0.7916	0.7782	0.7916	0.5712	0.5765	0.0191	0.0185	0.9744	0.9844	0.7916	0.5765	0.7916	0.5563	0.7805
0.4841	12.0	1929	0.6845	0.7978	0.7922	0.7978	0.6111	0.6226	0.0184	0.0178	0.9759	0.9849	0.7978	0.6226	0.7978	0.6092	0.7927
0.3838	13.0	2089	0.6929	0.7986	0.7947	0.7986	0.6347	0.6038	0.0184	0.0177	0.9743	0.9849	0.7986	0.6038	0.7986	0.5954	0.7903
0.3838	14.0	2250	0.6929	0.8017	0.7960	0.8017	0.6369	0.6270	0.0180	0.0174	0.9754	0.9851	0.8017	0.6270	0.8017	0.6174	0.7952
0.3838	14.93	2400	0.6841	0.8048	0.7955	0.8048	0.6332	0.6316	0.0177	0.0170	0.9753	0.9853	0.8048	0.6316	0.8048	0.6233	0.7978

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.1