phasen

phasen copied to clipboard

不知道作者有没有计划开源模型，让我们更方便的使用

yds1024

It throw the error that: operands could not be broadcast together with shapes (514,399) (257,397)

fmbao

Loss does not decrease

1

你好，感谢您的复现工作，不过我使用自己的数据训练该模型，loss不会下降，请问我该如何排查原因？ 我的数据为中英文均包含的干净录音，添加musan噪声后作为训练数据，使用mixloss，mixloss值稳定在40，sisnr值稳定在7～8之间，且不会下降和提升。

Wangzhen-kris

音频连接处有哒哒的声音或者消音的情况

3

你号，音频分成4秒每段进行语音增强后，在音频的连接处有哒哒的声音或者会出现消音的情况，将4s改成1s后的效果更加严重，这种情况可以采用什么方式去除呢？产生的原因是因为音频不连续吗？

SongJinXue

Mixloss 出现 nan

7

大佬，我用的是Mixloss，一运行loss就 nan. 1、LR 我已经设置很小了（0.00001）； 2、没有/0 情况；请问还有可能是什么原因呢？

jasdasdf

Loss fitting

5

想问下这个模型较好的拟合，loss值要接近多少，用的是-5-20信噪比的aishell数据，目前相位loss有点大

Chen1399

I am trying to reproduce the PHASEN, but I have a problem about data preprocessing. When the audio signal time is less than 4 seconds, what should I do? I...

HieDean

How training with cpu

2

I want to run this script, but my computer does not have a GPU. I tried to use the CPU to train, but it failed. How can it be compatible...

jay678

How to use tensorflow to conv_stft?

1

Hi,I use tensorflow to conv_stft like this: def init_kernels(win_len, win_inc, fft_len, win_type=None, invers=False): if win_type == 'None' or win_type is None: window = np.ones(win_len) else: window = get_window(win_type, win_len, fftbins=True)**0.5...

panhu

Fix Nan loss

2

I got "Nan" when use Mix loss to train (not speech denoise task), and Fix it by adding grad clip as fellows: loss.backward() nn.utils.clip_grad_norm_(self.estimator.parameters(), 10.0) # add this to clip...

makaichi