ruBert-base-finetuned-pos

This model was finetuned from ai-forever/ruBert-base on the disk0dancer/ru_sentances_pos dataset. All docs and code can be found on Github.

It achieves the following results on the evaluation set:

• eval_loss: 0.1544
• eval_precision: 0.8561
• eval_recall: 0.8723
• eval_f1: 0.8642
• eval_accuracy: 0.8822
• eval_runtime: 0.2476
• eval_samples_per_second: 80.775
• eval_steps_per_second: 8.078
• step: 0

Model description

Bert + Dence + Softmax + Dropout

Training and evaluation data

Model Trained for Token Classification

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Framework versions

• Transformers 4.39.0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2

Cite

@misc{churakov2024postagginghighlightskeletalstructure,
      title={POS-tagging to highlight the skeletal structure of sentences}, 
      author={Grigorii Churakov},
      year={2024},
      eprint={2411.14393},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.14393}, 
}