WhiteFu

使用新版本代码古诗生成问题

1

使用新版本代码，加载提供的古诗模型，想生成古诗的时候，发现使用generation.py得到的结果  第二句开始是四个字，后面的是五个字 但是用提供的huggingface的demo时  返回的是正常的五字诗，用的是同一个模型，为什么解码的格式不一样呢，明显huggingface才是诗的格式。 是哪里有问题吗，请指教

The effect of audio Split or not split?

1

Does audio segmentation(split) in data processing affect the model performance? Is it just for running the model to be able to support more batch size?

ERROR (speech-aligner[5.4.215~4-f2b7]:Input():util/kaldi-io.cc:756) Error opening input stream res/tree

4

I meet a error as follows: ERROR (speech-aligner[5.4.215~4-f2b7]:Input():util/kaldi-io.cc:756) Error opening input stream res/tree

How many gpus did you use during your training?

1

How many GPUs did you use during your training? I am training in a single GPU with batchsize = 16, and my result is weird.

hello, good job！ Can you provide some samples? thanks!

Loss exploded???

9

I get the error "loss explode" in the training stage! I'm not modifying the original hyperparameters, and I want to know how to solve the problem.

Good Job, but I found the pretrain model's line is unaviaiable.

Paper Link

1

Great job! Where can I get this paper？

Good Job! what's the repo envs setting?

关于libriTTS数据集

8

大佬，我在网上openslr上面下载的libriTTS 的clean100数据集和你们处理好的不一样，文本和语音都对不上，是我下错了还是需要额外的处理呢？求解惑