Bokai Xu

add chinese character and english words generation count evaluation (17% accuracy, 40 samples)

11

# Thank you for contributing an eval! ♥️ 🚨 Please make sure your PR follows these guidelines, __failure to follow the guidelines below will result in the PR being closed...

may i have a look at the training loss curve

3

Dear author: Thank you for your great work, i have used your code to train my model, but this is my first time to finetune a causal language model, i...

### Is there an existing issue ? / 是否已有相关的 issue ? - [X] I have searched, and there is no existing issue. / 我已经搜索过了，没有相关的 issue。 ### Describe the bug /...

s3tokenizer，g k开头的字容易有颤音

2

您好！用 cosyvoice2 25hz 的 s3tokenizer 尝试复现 cosyvoice2，训出来的模型音质挺好的，唯一美中不足的是 部分 g k 开头的字，比如 个、各、国、哥 这些字，有一定的几率会出现颤音，想问问作者这可能是啥原因呢，非常感谢！ 例如下面这个case： https://drive.google.com/file/d/1MUv8r1zhQ2Qoakg-9k4gPJHP1NaJnRVJ/view?usp=sharing 对应文本是 “日本将一个惊天秘密隐藏七十六年，全世界的人都被他给骗了。”

0.13.0 read wav failed

1

Got a bug when using soundfile==0.13.0 to load a wav byte Error calculating duration: Error opening : File contains data in an unknown format.