bert-large-cased-finetuned-ner-harem

This model is a fine-tuned version of google-bert/bert-large-cased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2665
• Precision: 0.7241
• Recall: 0.7423
• F1: 0.7331
• Accuracy: 0.9611

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.9938	140	0.2609	0.5107	0.5626	0.5354	0.9324
No log	1.9947	281	0.2057	0.6370	0.6642	0.6503	0.9517
No log	2.9956	422	0.2106	0.6527	0.6642	0.6584	0.9566
0.2074	3.9965	563	0.2342	0.6843	0.7054	0.6947	0.9571
0.2074	4.9973	704	0.2369	0.7216	0.7290	0.7253	0.9614
0.2074	5.9982	845	0.2334	0.7013	0.7261	0.7135	0.9574
0.2074	6.9991	986	0.2580	0.7139	0.7570	0.7348	0.9592
0.0377	8.0	1127	0.2658	0.7452	0.7452	0.7452	0.9607
0.0377	8.9938	1267	0.2619	0.7543	0.7688	0.7615	0.9637
0.0377	9.9379	1400	0.2665	0.7241	0.7423	0.7331	0.9611

Framework versions

• Transformers 4.41.1
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1