metadata

license: mit
tags:
  - generated_from_trainer
base_model: law-ai/InLegalBERT
metrics:
  - accuracy
  - precision
  - recall
model-index:
  - name: InLegalBERT
    results: []

Metrics

loss: 1.0342
accuracy: 0.8359
precision: 0.8409
recall: 0.8359
precision_macro: 0.8136
recall_macro: 0.8000
macro_fpr: 0.0142
weighted_fpr: 0.0138
weighted_specificity: 0.9792
macro_specificity: 0.9877
weighted_sensitivity: 0.8359
macro_sensitivity: 0.8000
f1_micro: 0.8359
f1_macro: 0.8010
f1_weighted: 0.8352
runtime: 19.9583
samples_per_second: 64.4340
steps_per_second: 8.0670

Metrics

loss: 1.0345
accuracy: 0.8358
precision: 0.8408
recall: 0.8358
precision_macro: 0.8207
recall_macro: 0.7957
macro_fpr: 0.0143
weighted_fpr: 0.0138
weighted_specificity: 0.9790
macro_specificity: 0.9877
weighted_sensitivity: 0.8358
macro_sensitivity: 0.7957
f1_micro: 0.8358
f1_macro: 0.8020
f1_weighted: 0.8352
runtime: 22.0569
samples_per_second: 58.5300
steps_per_second: 7.3450

InLegalBERT

This model is a fine-tuned version of law-ai/InLegalBERT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.0763
Accuracy: 0.8304
Precision: 0.8363
Recall: 0.8304
Precision Macro: 0.7959
Recall Macro: 0.8029
Macro Fpr: 0.0150
Weighted Fpr: 0.0145
Weighted Specificity: 0.9774
Macro Specificity: 0.9871
Weighted Sensitivity: 0.8296
Macro Sensitivity: 0.8029
F1 Micro: 0.8296
F1 Macro: 0.7954
F1 Weighted: 0.8283

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
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 15
mixed_precision_training: Native AMP

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
1.065	1.0	643	0.6395	0.7994	0.7818	0.7994	0.6194	0.6308	0.0185	0.0176	0.9714	0.9847	0.7994	0.6308	0.7994	0.6029	0.7804
0.5866	2.0	1286	0.6907	0.8187	0.8199	0.8187	0.7285	0.7366	0.0161	0.0156	0.9765	0.9864	0.8187	0.7366	0.8187	0.7276	0.8152
0.4622	3.0	1929	0.8056	0.8180	0.8137	0.8180	0.7227	0.7376	0.0162	0.0156	0.9764	0.9863	0.8180	0.7376	0.8180	0.7283	0.8150
0.2398	4.0	2572	0.9310	0.8172	0.8235	0.8172	0.7661	0.7425	0.0161	0.0157	0.9762	0.9862	0.8172	0.7425	0.8172	0.7407	0.8161
0.1611	5.0	3215	1.0763	0.8304	0.8363	0.8304	0.8174	0.7918	0.0148	0.0144	0.9784	0.9873	0.8304	0.7918	0.8304	0.7986	0.8304
0.1055	6.0	3858	1.1377	0.8257	0.8275	0.8257	0.8039	0.7810	0.0154	0.0149	0.9775	0.9869	0.8257	0.7810	0.8257	0.7863	0.8246
0.0463	7.0	4501	1.3215	0.8071	0.8111	0.8071	0.7692	0.7689	0.0172	0.0168	0.9761	0.9856	0.8071	0.7689	0.8071	0.7661	0.8078
0.031	8.0	5144	1.3483	0.8203	0.8170	0.8203	0.7773	0.7727	0.0161	0.0154	0.9751	0.9864	0.8203	0.7727	0.8203	0.7690	0.8175
0.0202	9.0	5787	1.3730	0.8280	0.8263	0.8280	0.7818	0.7803	0.0152	0.0146	0.9779	0.9871	0.8280	0.7803	0.8280	0.7753	0.8256
0.0133	10.0	6430	1.5407	0.8164	0.8163	0.8164	0.7688	0.7779	0.0165	0.0158	0.9751	0.9861	0.8164	0.7779	0.8164	0.7655	0.8135
0.0051	11.0	7073	1.5235	0.8226	0.8265	0.8226	0.7900	0.7680	0.0156	0.0152	0.9769	0.9866	0.8226	0.7680	0.8226	0.7744	0.8234
0.0027	12.0	7716	1.5643	0.8265	0.8259	0.8265	0.7805	0.7841	0.0154	0.0148	0.9772	0.9869	0.8265	0.7841	0.8265	0.7775	0.8245
0.002	13.0	8359	1.5516	0.8280	0.8273	0.8280	0.7882	0.7902	0.0152	0.0146	0.9779	0.9871	0.8280	0.7902	0.8280	0.7860	0.8262
0.0015	14.0	9002	1.5835	0.8273	0.8268	0.8273	0.7943	0.8022	0.0153	0.0147	0.9773	0.9870	0.8273	0.8022	0.8273	0.7943	0.8259
0.0007	15.0	9645	1.5914	0.8296	0.8293	0.8296	0.7959	0.8029	0.0150	0.0145	0.9774	0.9871	0.8296	0.8029	0.8296	0.7954	0.8283

Framework versions

Transformers 4.38.2
Pytorch 2.1.2
Datasets 2.1.0
Tokenizers 0.15.2