|
--- |
|
language: ja |
|
thumbnail: https://github.com/rinnakk/japanese-gpt2/blob/master/rinna.png |
|
tags: |
|
- ja |
|
- japanese |
|
- gpt2 |
|
- text-generation |
|
- lm |
|
- nlp |
|
license: mit |
|
datasets: |
|
- cc100 |
|
- wikipedia |
|
--- |
|
|
|
# japanese-gpt2-xsmall |
|
|
|
 |
|
|
|
This repository provides an extra-small-sized Japanese GPT-2 model. The model is provided by [rinna](https://corp.rinna.co.jp/). |
|
|
|
# How to use the model |
|
|
|
*NOTE:* Use `T5Tokenizer` to initiate the tokenizer. |
|
|
|
~~~~ |
|
from transformers import T5Tokenizer, GPT2LMHeadModel |
|
|
|
tokenizer = T5Tokenizer.from_pretrained("rinna/japanese-gpt2-small") |
|
tokenizer.do_lower_case = True # due to some bug of tokenizer config loading |
|
|
|
model = GPT2LMHeadModel.from_pretrained("rinna/japanese-gpt2-small") |
|
~~~~ |
|
|
|
# Model architecture |
|
A 6-layer, 512-hidden-size transformer-based language model. |
|
|
|
# Training |
|
The model was trained on [Japanese CC-100](http://data.statmt.org/cc-100/ja.txt.xz) and [Japanese Wikipedia](https://dumps.wikimedia.org/other/cirrussearch) to optimize a traditional language modelling objective on 8\\*V100 GPUs for around 4 days. It reaches around 28 perplexity on a chosen validation set from CC-100. |
|
|
|
# Tokenization |
|
The model uses a [sentencepiece](https://github.com/google/sentencepiece)-based tokenizer, the vocabulary was trained on the Japanese Wikipedia using the official sentencepiece training script. |
|
|
|
# Licenese |
|
[The MIT license](https://opensource.org/licenses/MIT) |
|
|
