Commit
•
6a9922d
1
Parent(s):
470190e
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: pl
|
| 3 |
+
tags:
|
| 4 |
+
- T5
|
| 5 |
+
- translation
|
| 6 |
+
- summarization
|
| 7 |
+
- question answering
|
| 8 |
+
- reading comprehension
|
| 9 |
+
datasets:
|
| 10 |
+
- ccnet
|
| 11 |
+
- nkjp
|
| 12 |
+
- wikipedia
|
| 13 |
+
- open subtitles
|
| 14 |
+
- free readings
|
| 15 |
+
license: cc-by-4.0
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# plT5 Large
|
| 19 |
+
**plT5** models are T5-based language models trained on Polish corpora. Models were optimized for the original T5 denoising target.
|
| 20 |
+
|
| 21 |
+
## Corpus
|
| 22 |
+
plT5 was trained on six different corpora available for Polish language:
|
| 23 |
+
|
| 24 |
+
| Corpus | Tokens | Documents |
|
| 25 |
+
| :------ | ------: | ------: |
|
| 26 |
+
| [CCNet Middle](https://github.com/facebookresearch/cc_net) | 3243M | 7.9M |
|
| 27 |
+
| [CCNet Head](https://github.com/facebookresearch/cc_net) | 2641M | 7.0M |
|
| 28 |
+
| [National Corpus of Polish](http://nkjp.pl/index.php?page=14&lang=1)| 1357M | 3.9M |
|
| 29 |
+
| [Open Subtitles](http://opus.nlpl.eu/OpenSubtitles-v2018.php) | 1056M | 1.1M
|
| 30 |
+
| [Wikipedia](https://dumps.wikimedia.org/) | 260M | 1.4M |
|
| 31 |
+
| [Wolne Lektury](https://wolnelektury.pl/) | 41M | 5.5k |
|
| 32 |
+
|
| 33 |
+
## Tokenizer
|
| 34 |
+
The training dataset was tokenized into subwords using a sentencepiece unigram with
|
| 35 |
+
vocabulary size of 50k tokens.
|
| 36 |
+
|
| 37 |
+
## Usage
|
| 38 |
+
Example code:
|
| 39 |
+
```python
|
| 40 |
+
from transformers import AutoTokenizer, AutoModel
|
| 41 |
+
|
| 42 |
+
tokenizer = AutoTokenizer.from_pretrained("allegro/plT5-large")
|
| 43 |
+
model = AutoModel.from_pretrained("allegro/plT5-large")
|
| 44 |
+
```
|
| 45 |
+
|
| 46 |
+
## License
|
| 47 |
+
CC BY 4.0
|
| 48 |
+
|
| 49 |
+
## Citation
|
| 50 |
+
If you use this model, please cite the following paper:
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
## Authors
|
| 56 |
+
The model was trained by [**Machine Learning Research Team at Allegro**](https://ml.allegro.tech/) and [**Linguistic Engineering Group at Institute of Computer Science, Polish Academy of Sciences**](http://zil.ipipan.waw.pl/).
|
| 57 |
+
|
| 58 |
+
You can contact us at: <a href="mailto:klejbenchmark@allegro.pl">klejbenchmark@allegro.pl</a>
|