distillbert-base-cased-finetuned-ner6

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

• Loss: 0.6861
• Precision: 0.8244
• Recall: 0.8485
• F1: 0.8363
• Accuracy: 0.9634

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: 32
• eval_batch_size: 32
• seed: 5226
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 13
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.7373	1.0	1000	0.7242	0.7652	0.7968	0.7807	0.9523
0.6986	2.0	2000	0.6944	0.7920	0.8249	0.8081	0.9581
0.6785	3.0	3000	0.6893	0.8026	0.8291	0.8156	0.9598
0.6749	4.0	4000	0.6847	0.7984	0.8352	0.8164	0.9601
0.6626	5.0	5000	0.6826	0.8121	0.8373	0.8245	0.9617
0.6637	6.0	6000	0.6807	0.8176	0.8415	0.8294	0.9623
0.6528	7.0	7000	0.6812	0.8134	0.8425	0.8277	0.9626
0.6532	8.0	8000	0.6816	0.8209	0.8453	0.8329	0.9629
0.6431	9.0	9000	0.6814	0.8178	0.8447	0.8311	0.9631
0.6444	10.0	10000	0.6801	0.8176	0.8466	0.8318	0.9629
0.6347	11.0	11000	0.6826	0.8214	0.8470	0.8340	0.9631
0.6377	12.0	12000	0.6849	0.8214	0.8468	0.8339	0.9630
0.628	13.0	13000	0.6861	0.8244	0.8485	0.8363	0.9634

Framework versions

• Transformers 4.50.1
• Pytorch 2.5.1+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1