turkic-cyrillic-classifier

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

{'test_loss': 0.013604652136564255,
 'test_accuracy': 0.997,
 'test_f1': 0.9969996069718668,
 'test_runtime': 60.5479,
 'test_samples_per_second': 148.643,
 'test_steps_per_second': 2.329}

Model description

The model classifies text based on a provided Turkic language written in Cyrillic script.

Languages

• bak - Bashkir
• chv - Chuvash
• sah - Sakha
• tat - Tatar
• kir - Kyrgyz
• kaz - Kazakh
• tyv - Tuvinian
• krc - Karachay-Balkar
• rus - Russian

Intended uses & limitations

Training and evaluation data

cyrillic_turkic_langs

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.1063	1.0	1000	0.0204	0.9950	0.9950
0.0126	2.0	2000	0.0136	0.9970	0.9970

Framework versions

• Transformers 4.27.0
• Pytorch 1.13.1+cu116
• Datasets 2.10.1