bert-large-uncased-finetuned-ner-harem

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

• Loss: 0.3109
• Precision: 0.6895
• Recall: 0.6442
• F1: 0.6661
• Accuracy: 0.9512

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.9978	281	0.2896	0.5442	0.4772	0.5085	0.9238
0.3496	1.9973	562	0.2340	0.6811	0.5295	0.5958	0.9412
0.3496	2.9969	843	0.2240	0.5876	0.5599	0.5734	0.9409
0.1372	3.9964	1124	0.2540	0.6910	0.6223	0.6548	0.9403
0.1372	4.9960	1405	0.2598	0.6433	0.6358	0.6395	0.9439
0.0648	5.9956	1686	0.2377	0.6945	0.6442	0.6684	0.9497
0.0648	6.9951	1967	0.2822	0.6965	0.6425	0.6684	0.9501
0.0316	7.9982	2249	0.2958	0.7044	0.6509	0.6766	0.9518
0.0148	8.9978	2530	0.3006	0.6944	0.6476	0.6702	0.9496
0.0148	9.9938	2810	0.3109	0.6895	0.6442	0.6661	0.9512

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3