---
license: mit
base_model: FacebookAI/xlm-roberta-large
tags:
- generated_from_trainer
model-index:
- name: xlm-roberta-large-ft-qa-ru-mt-to-kaz
  results: []
datasets:
- med-alex/qa_mt_ru_to_kaz
language:
- kk
metrics:
- exact_match
- f1
library_name: transformers
pipeline_tag: question-answering
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# xlm-roberta-large-ft-qa-ru-mt-to-kaz

This model is a fine-tuned version of [FacebookAI/xlm-roberta-large](https://huggingface.co/FacebookAI/xlm-roberta-large) on the med-alex/qa_mt_ru_to_kaz dataset.

## Model description

This model is one of many models created within the framework of a project to study the solution of a QA task for low-resource languages using the example of Kazakh and Uzbek. 

Please see the [description](https://github.com/med-alex/turkic_qa?tab=readme-ov-file#добро-пожаловать-на-студенческий-проект-посвященный-решению-задачи-qa-для-низкоресурсных-языков-на-примере-казахского-и-узбекского-языка) of the project, where there is a description of the solution and the results of the models in order to choose the best model for the Kazakh or Uzbek language.

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 28
- eval_batch_size: 28
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 5.0

### Framework versions

- Transformers 4.40.1
- Pytorch 2.0.0+cu118
- Datasets 2.18.0
- Tokenizers 0.19.1