Hugging Face's logo

megagonlabs
/
t5-base-japanese-web-8k

Text2Text Generation
Transformers
PyTorch
Japanese
t5
seq2seq
text-generation-inference
Inference Endpoints
Model card Files Community
2
t5-base-japanese-web-8k
File size: 2,626 Bytes 
9b69d11
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
---
language: ja
tags:
- t5
- text2text-generation
- seq2seq
license: apache-2.0
datasets:
- mc4
- wiki40b
---

# t5-base-japanese-web (with Byte-fallback, 8K)

## Description

[megagonlabs/t5-base-japanese-web](https://huggingface.co/megagonlabs/t5-base-japanese-web) is a T5 (Text-to-Text Transfer Transformer) model pre-trained on Japanese web texts.  
Training codes are [available on GitHub](https://github.com/megagonlabs/t5-japanese).

The vocabulary size of this model is 8K.
[32K version is also available](https://huggingface.co/megagonlabs/t5-base-japanese-web).

### Corpora

We used following corpora for pre-training.

- Japanese in [mC4/3.0.1](https://huggingface.co/datasets/mc4) (We used [Tensorflow native format](https://github.com/allenai/allennlp/discussions/5056))
    - 87,425,304 pages
    - 782 GB in TFRecord format
- [Japanese](https://www.tensorflow.org/datasets/catalog/wiki40b#wiki40bja) in [wiki40b/1.3.0](https://www.tensorflow.org/datasets/catalog/wiki40b)
    - 828,236 articles (2,073,584 examples)
    - 2 GB in TFRecord format

### Tokenizer

We used Japanese Wikipedia to train [SentencePiece](https://github.com/google/sentencepiece).

- Vocabulary size: 8,000
- [Byte-fallback](https://github.com/google/sentencepiece/releases/tag/v0.1.9): Enabled

### Parameters

- T5 model: [models/t5.1.1.base.gin](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/t5/models/gin/models/t5.1.1.base.gin)
- Training steps: 1,000,000

It took about 126 hours with TPU v3-8

## Related models

- [日本語T5事前学習済みモデル (sonoisa/t5-base-japanese)](https://huggingface.co/sonoisa/t5-base-japanese)
- [日本語T5事前学習済みモデル (sonoisa/t5-base-japanese-mC4-Wikipedia)](https://huggingface.co/sonoisa/t5-base-japanese-mC4-Wikipedia)

## License

Apache License 2.0

## Citations

- mC4

Contains information from `mC4` which is made available under the [ODC Attribution License](https://opendatacommons.org/licenses/by/1-0/).

```bibtex
@article{2019t5,
    author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
    title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
    journal = {arXiv e-prints},
    year = {2019},
    archivePrefix = {arXiv},
    eprint = {1910.10683},
}
```

- wiki40b

```bibtex
@inproceedings{49029,
title = {Wiki-40B: Multilingual Language Model Dataset},
author = {Mandy Guo and Zihang Dai and Denny Vrandecic and Rami Al-Rfou},
year = {2020},
booktitle   = {LREC 2020}
}
```