ukrainian-qa

This model is a fine-tuned version of ukr-models/xlm-roberta-base-uk on the UA-SQuAD dataset.

Link to training scripts - https://github.com/robinhad/ukrainian-qa
It achieves the following results on the evaluation set:

• Loss: 1.4778

Model description

More information needed

How to use

from transformers import pipeline, AutoTokenizer, AutoModelForQuestionAnswering
model_name = "robinhad/ukrainian-qa"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForQuestionAnswering.from_pretrained(model_name)

qa_model = pipeline("question-answering", model=model.to("cpu"), tokenizer=tokenizer)
question = "Де ти живеш?"
context = "Мене звати Сара і я живу у Лондоні"
qa_model(question = question, context = context)

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: 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: 6

Training results

Training Loss	Epoch	Step	Validation Loss
2.4526	1.0	650	1.3631
1.3317	2.0	1300	1.2229
1.0693	3.0	1950	1.2184
0.6851	4.0	2600	1.3171
0.5594	5.0	3250	1.3893
0.4954	6.0	3900	1.4778

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0
• Datasets 2.2.2
• Tokenizers 0.12.1