Flavio

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: 0.3914
• Accuracy: 0.9150
• F1 Macro: 0.8231
• F1 Class 0: 0.9472
• F1 Class 1: 0.6667
• F1 Class 2: 0.9259
• F1 Class 3: 0.8421
• F1 Class 4: 0.9
• F1 Class 5: 0.9615
• F1 Class 6: 0.8
• F1 Class 7: 0.9556
• F1 Class 8: 0.9655
• F1 Class 9: 0.8621
• F1 Class 10: 0.8924
• F1 Class 11: 0.7143
• F1 Class 12: 0.8101
• F1 Class 13: 0.75
• F1 Class 14: 0.8889
• F1 Class 15: 0.7500
• F1 Class 16: 0.0
• F1 Class 17: 0.9880
• F1 Class 18: 0.9180
• F1 Class 19: 0.9231

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: 2e-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
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 Class 0	F1 Class 1	F1 Class 2	F1 Class 3	F1 Class 4	F1 Class 5	F1 Class 6	F1 Class 7	F1 Class 8	F1 Class 9	F1 Class 10	F1 Class 11	F1 Class 12	F1 Class 13	F1 Class 14	F1 Class 15	F1 Class 16	F1 Class 17	F1 Class 18	F1 Class 19
1.2343	0.39	250	0.7445	0.8363	0.5500	0.875	0.0	0.8959	0.8421	0.0769	0.6818	0.6667	0.9556	0.9492	0.6190	0.8339	0.0	0.7442	0.2000	0.8267	0.0	0.0	0.9760	0.8571	0.0
0.5654	0.79	500	0.5466	0.8690	0.6846	0.9124	0.0	0.9189	0.8421	0.7660	0.8302	0.6531	0.9663	0.9310	0.7353	0.8580	0.0	0.7564	0.8889	0.8272	0.1	0.0	0.9759	0.8070	0.9231
0.439	1.18	750	0.4626	0.8832	0.7211	0.9209	0.0	0.9217	0.8421	0.8	0.9057	0.6667	0.9556	0.9455	0.8000	0.8554	0.2857	0.7799	0.8889	0.8462	0.1905	0.0	0.9759	0.9180	0.9231
0.3397	1.57	1000	0.4744	0.8885	0.7457	0.9207	0.0	0.9327	0.8421	0.7826	0.8364	0.7547	0.9663	0.9655	0.7273	0.8735	0.6667	0.8077	0.8889	0.8553	0.32	0.0	0.9730	0.8772	0.9231
0.3351	1.97	1250	0.4128	0.8938	0.7784	0.9350	0.4	0.9217	0.8000	0.8108	0.8519	0.6939	0.9663	0.9474	0.7719	0.8563	0.7692	0.8199	0.8889	0.8903	0.4800	0.0	0.9790	0.8621	0.9231
0.2384	2.36	1500	0.3982	0.9071	0.8016	0.9431	0.4	0.9259	0.8421	0.9048	0.8772	0.8333	0.9556	0.9655	0.8302	0.8810	0.6667	0.7922	0.8889	0.8961	0.5882	0.0	0.9850	0.9333	0.9231
0.2309	2.75	1750	0.3741	0.9133	0.8191	0.9494	0.6667	0.9266	0.8421	0.8780	0.9091	0.8197	0.9556	0.9655	0.84	0.8831	0.625	0.8026	0.8235	0.9032	0.7647	0.0	0.9880	0.9153	0.9231
0.2243	3.14	2000	0.3962	0.9080	0.8146	0.9435	0.5714	0.9302	0.8421	0.9	0.9804	0.7059	0.9556	0.9492	0.8727	0.8765	0.7692	0.8050	0.8235	0.8889	0.6452	0.0	0.9760	0.9333	0.9231
0.1781	3.54	2250	0.3775	0.9133	0.8137	0.9418	0.4	0.9395	0.8421	0.9	0.9091	0.8814	0.9556	0.9655	0.8421	0.8952	0.7143	0.8077	0.8235	0.8679	0.7500	0.0	0.9816	0.9333	0.9231
0.169	3.93	2500	0.4092	0.9080	0.8157	0.9395	0.6667	0.9224	0.8421	0.9	0.9091	0.8136	0.9556	0.9655	0.8621	0.8825	0.6667	0.8077	0.75	0.8701	0.7500	0.0	0.9879	0.9	0.9231
0.1406	4.32	2750	0.3886	0.9097	0.8244	0.9424	0.5714	0.9266	0.8421	0.9048	0.9615	0.7931	0.9556	0.9492	0.8667	0.8790	0.7692	0.7949	0.8889	0.8718	0.7273	0.0	0.9849	0.9355	0.9231
0.1245	4.72	3000	0.3914	0.9150	0.8231	0.9472	0.6667	0.9259	0.8421	0.9	0.9615	0.8	0.9556	0.9655	0.8621	0.8924	0.7143	0.8101	0.75	0.8889	0.7500	0.0	0.9880	0.9180	0.9231

Framework versions

• Transformers 4.32.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.4
• Tokenizers 0.13.3