File size: 2,379 Bytes
43ba99f 2abacad 9cefb87 43ba99f 2abacad 74057b1 2abacad 9cefb87 2abacad 9cefb87 74057b1 2abacad 74057b1 954ede7 2abacad 74057b1 2abacad 9cefb87 |
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 |
---
language: ja
tags:
- speech
license: other
---
# distilhubert-ft-japanese-50k
Fine-tuned (more precisely, continue trained) 50k steps model on Japanese using the [JVS corpus](https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus), [Tsukuyomi-Chan corpus](https://tyc.rei-yumesaki.net/material/corpus/), [Amitaro's ITA corpus V2.1](https://amitaro.net/), and recorded my own read [ITA corpus](https://github.com/mmorise/ita-corpus).
## Attention
This checkpoint was used the [JVS corpus](https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus) when training. Please read and accept the [terms of use](https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus#h.p_OP_G8FT_Kuf4)
(This terms of use was also applies this checkpoint. This means also applies this "terms of use" when you use this checkpoint with another project etc...)
# References
Original repos, Many thanks!:
[S3PRL](https://github.com/s3prl/s3prl/tree/main/s3prl/pretrain)
- Using this when training (with little modify for train using own datasets).
[distilhubert (hf)](https://huggingface.co/ntu-spml/distilhubert)
Note: As same as the original, this model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
# Usage
See [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more information on how to fine-tune the model. Note that the class `Wav2Vec2ForCTC` has to be replaced by `HubertForCTC`.
Note: This is not the best checkpoint and become more accurate with continued train, I think. I'll try to continue when I have a time.
## Credits
- [JVS corpus](https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus)
- [Tsukuyomi-Chan corpus](https://tyc.rei-yumesaki.net/material/corpus/)
```
■つくよみちゃんコーパス（CV.夢前黎）
```
[https://tyc.rei-yumesaki.net/material/corpus/](https://tyc.rei-yumesaki.net/material/corpus/)
- [Amitaro's ITA corpus](https://amitaro.net/)
```
あみたろの声素材工房
```
[https://amitaro.net/](https://amitaro.net/)
Thanks! |