legal_deberta

tthhanh/legal_deberta

legal_deberta is a machine learning model from tthhanh. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

3% of all-time downloads

All-time downloads

349

Public

Repo size

39.7 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin735 MB · 100%

At a glance

License: mit
Model type: deberta-v2
Access: Public
Created: Aug 12, 2024
Updated: Aug 12, 2024
SHA: 9a6e984f

Base models

microsoft/deberta-v3-base

Model card

From the Hugging Face model README

legal_deberta

This model is a fine-tuned version of microsoft/deberta-v3-base on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.2674
Law Precision: 0.6932
Law Recall: 0.8133
Law F1: 0.7485
Law Number: 75
Violated by Precision: 0.8684
Violated by Recall: 0.88
Violated by F1: 0.8742
Violated by Number: 75
Violated on Precision: 0.5882
Violated on Recall: 0.6667
Violated on F1: 0.625
Violated on Number: 75
Violation Precision: 0.5287
Violation Recall: 0.6429
Violation F1: 0.5802
Violation Number: 616
Overall Precision: 0.5741
Overall Recall: 0.6813
Overall F1: 0.6232
Overall Accuracy: 0.9461

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: 500
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Law Precision	Law Recall	Law F1	Law Number	Violated by Precision	Violated by Recall	Violated by F1	Violated by Number	Violated on Precision	Violated on Recall	Violated on F1	Violated on Number	Violation Precision	Violation Recall	Violation F1	Violation Number	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
1.9748	1.0	45	1.1555	0.0	0.0	0.0	75	0.0	0.0	0.0	75	0.0	0.0	0.0	75	0.0	0.0	0.0	616	0.0	0.0	0.0	0.7437
0.4536	2.0	90	0.3670	0.0	0.0	0.0	75	0.0	0.0	0.0	75	0.0	0.0	0.0	75	0.1704	0.2955	0.2162	616	0.1704	0.2164	0.1907	0.8901
0.2704	3.0	135	0.2199	0.7059	0.64	0.6713	75	0.3095	0.1733	0.2222	75	0.0909	0.0133	0.0233	75	0.3291	0.5097	0.4000	616	0.3498	0.4471	0.3925	0.9277
0.1475	4.0	180	0.1959	0.6263	0.8267	0.7126	75	0.9153	0.72	0.8060	75	0.3182	0.3733	0.3436	75	0.4641	0.5974	0.5224	616	0.4928	0.6088	0.5447	0.9407
0.0879	5.0	225	0.2038	0.5909	0.8667	0.7027	75	0.7590	0.84	0.7975	75	0.3982	0.6	0.4787	75	0.4692	0.6055	0.5287	616	0.4959	0.6492	0.5623	0.9434
0.0499	6.0	270	0.2466	0.5913	0.9067	0.7158	75	0.7674	0.88	0.8199	75	0.4412	0.6	0.5085	75	0.4832	0.6071	0.5381	616	0.5135	0.6576	0.5766	0.9425
0.0291	7.0	315	0.2980	0.5755	0.8133	0.6740	75	0.7976	0.8933	0.8428	75	0.3802	0.6133	0.4694	75	0.4929	0.5617	0.5250	616	0.5133	0.6183	0.5609	0.9389
0.0341	8.0	360	0.2660	0.5739	0.88	0.6947	75	0.8193	0.9067	0.8608	75	0.48	0.64	0.5486	75	0.4800	0.6445	0.5502	616	0.5147	0.6885	0.5890	0.9366
0.0228	9.0	405	0.3186	0.3505	0.9067	0.5056	75	0.6126	0.9067	0.7312	75	0.3216	0.7333	0.4472	75	0.4365	0.5519	0.4875	616	0.4231	0.6314	0.5067	0.9301
0.0173	10.0	450	0.2674	0.6932	0.8133	0.7485	75	0.8684	0.88	0.8742	75	0.5882	0.6667	0.625	75	0.5287	0.6429	0.5802	616	0.5741	0.6813	0.6232	0.9461

Framework versions

Transformers 4.31.0
Pytorch 2.4.0+cu121
Datasets 2.15.0
Tokenizers 0.13.3

Type: deberta-v2
License: mit
Created: Aug 12, 2024
Updated: Aug 12, 2024