privacy-300k-masking

This model is a fine-tuned version of distilbert-base-multilingual-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3729
• Overall Precision: 0.2655
• Overall Recall: 0.1856
• Overall F1: 0.2185
• Overall Accuracy: 0.8664
• Bod F1: 0.2060
• Building F1: 0.2527
• Cardissuer F1: 0.0
• City F1: 0.2253
• Country F1: 0.2800
• Date F1: 0.2289
• Driverlicense F1: 0.1902
• Email F1: 0.2350
• Geocoord F1: 0.1572
• Givenname1 F1: 0.2029
• Givenname2 F1: 0.1330
• Idcard F1: 0.2208
• Ip F1: 0.1826
• Lastname1 F1: 0.1877
• Lastname2 F1: 0.0937
• Lastname3 F1: 0.0328
• Pass F1: 0.1950
• Passport F1: 0.2256
• Postcode F1: 0.2518
• Secaddress F1: 0.2101
• Sex F1: 0.2636
• Socialnumber F1: 0.1891
• State F1: 0.2639
• Street F1: 0.1915
• Tel F1: 0.2077
• Time F1: 0.2551
• Title F1: 0.2453
• Username F1: 0.2325

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Overall Precision	Overall Recall	Overall F1	Overall Accuracy	Bod F1	Building F1	Cardissuer F1	City F1	Country F1	Date F1	Driverlicense F1	Email F1	Geocoord F1	Givenname1 F1	Givenname2 F1	Idcard F1	Ip F1	Lastname1 F1	Lastname2 F1	Lastname3 F1	Pass F1	Passport F1	Postcode F1	Secaddress F1	Sex F1	Socialnumber F1	State F1	Street F1	Tel F1	Time F1	Title F1	Username F1
0.3954	1.0	88839	0.3729	0.2655	0.1856	0.2185	0.8664	0.2060	0.2527	0.0	0.2253	0.2800	0.2289	0.1902	0.2350	0.1572	0.2029	0.1330	0.2208	0.1826	0.1877	0.0937	0.0328	0.1950	0.2256	0.2518	0.2101	0.2636	0.1891	0.2639	0.1915	0.2077	0.2551	0.2453	0.2325

Framework versions

• Transformers 4.40.0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1