Eddy Ma（马远洋）

got this error when `npm run watch`, any thought? 

I am using dotnet to generate report files in `coverage`, a sub folder with the naming rule of GUID is created for each `dotnet version` run, looks like codecov-action does...

Looking for the best practice for face recognition over thousand of people

1

Hi We have a case where we would like to do face recognition for over 1000 different employees, the current model's accuracy does not perform very well, any best practice...

wrong prediction result after convertion

6

### Step 1 - Training ``` from numpy import asarray from xgboost import XGBRegressor from sklearn.model_selection import cross_val_score from sklearn.model_selection import RepeatedKFold model = XGBRegressor(n_estimators=200, max_depth=7, eta=0.01, subsample=0.7, colsample_bytree=0.8) #...

ImportError: cannot import name 'parse' from 'basicsr.utils.options'

2

Getting the following error, installed by `!pip3 install basicsr` on colab, any specific version required? ImportError: cannot import name 'parse' from 'basicsr.utils.options'

Cantonese finetune效果比较差

4

尝试用tts_finetune的模式去做广东话克隆，声音是像的，可是就是电流声大，训练数据是来自于自己的麦克风录音，训练数据听起来很清晰的，可是finetune出来的结果就是大“震音/电流音”， 尝试用其他TTS生成的wav和差不多的量作为训练题材，克隆出来的效果很不错。请问效果不好是因为录音问题吗？ Finetune的步骤 1. 用MFA对齐 2. 使用了预训练模型fastspeech2_canton_ckpt_1.4.0.zip 和 pwg_aishell3_ckpt_0.5.zip 这个是训练出来的样本 [170.wav.zip](https://github.com/PaddlePaddle/PaddleSpeech/files/11591608/170.wav.zip)