Lagadro/teknofest_ner_mammo

This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on an private mammogram reports dataset. It achieves the following results on the evaluation set:

• Train Loss: 0.0812
• Validation Loss: 0.0961
• Epoch: 4
• Overall Accuracy: 0.968

Model description

This model is trained to extract entity names from reports of mammographic images.

Training and evaluation data

The data has been provided by Teknofest.

Training hyperparameters

• Name: AdamWeightDecay
• Learning Rate:
  • Module: keras.optimizers.schedules
  • Class: PolynomialDecay
  • Config:
    • Initial Learning Rate: 2e-05
    • Decay Steps: 260
    • End Learning Rate: 0.0
    • Power: 1.0
    • Cycle: False
    • Name: None
  • Registered Name: None
• Decay: 0.0
• Beta 1: 0.9
• Beta 2: 0.999
• Epsilon: 1e-08
• Amsgrad: False
• Weight Decay Rate: 0.01
• Training Precision: float32

Training results

Train Loss	Validation Loss	Epoch
0.5769	0.2226	0
0.1650	0.1344	1
0.1119	0.1110	2
0.0907	0.0994	3
0.0812	0.0961	4

Framework versions

• Transformers 4.41.2
• TensorFlow 2.15.0
• Datasets 2.19.2
• Tokenizers 0.19.1