File size: 3,127 Bytes
7aaa871
 
86bea78
a7c8bfb
 
 
0e66f0e
a7c8bfb
0e66f0e
 
 
 
7aaa871
a7c8bfb
 
 
 
0e66f0e
a7c8bfb
0e66f0e
4f3e0de
e0caec7
a7c8bfb
 
 
 
 
 
 
0e66f0e
 
c6e073a
0e66f0e
 
 
 
 
 
 
 
 
 
 
 
a7c8bfb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
---
license: mit
language: uk
tags:
- generated_from_trainer
model-index:
- name: ukrainian-qa
  results: []
widget:
- text: "Що відправлять для ЗСУ?"
  context: "Про це повідомив міністр оборони Арвідас Анушаускас. Уряд Литви не має наміру зупинятися у військово-технічній допомозі Україні. Збройні сили отримають антидрони, тепловізори та ударний безпілотник. «Незабаром Литва передасть Україні не лише обіцяні бронетехніку, вантажівки та позашляховики, але також нову партію антидронів та тепловізорів. І, звичайно, Байрактар, який придбають на зібрані литовцями гроші», - написав глава Міноборони."

---

<!-- 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. -->

# ukrainian-qa

This model is a fine-tuned version of [ukr-models/xlm-roberta-base-uk](https://huggingface.co/ukr-models/xlm-roberta-base-uk) on the [UA-SQuAD](https://github.com/fido-ai/ua-datasets/tree/main/ua_datasets/src/question_answering) dataset.

Link to training scripts - [https://github.com/robinhad/ukrainian-qa](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

```python
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