FastDiff
FastDiff copied to clipboard
Finetune on my owndataset
Hello, I want to finetune the model with my own dataset I want to understand more about how data structure should be in the following folder : raw_data_dir, processed_data_dir, binary_data_dir
Thanks
Hi, only the raw_data_dir needs to follow a specific structure, and the other two would be generated automatically. For the raw_data_dir, you could follow the code which organizes waveforms in pre_align.py, and I hope it could be helpful.
okay I've set the directory of raw data in the /content/FastDiff/modules/FastDiff/config/FastDiff.yaml file and it contains data of .wav format then I tried to run /content/FastDiff/align_and_binarize.py and I got the following error I don't know what should the hparamter be set ?
and wen I try to run binarize.py
I got this error message there's a file missing in utils folder
@abdelrahmanhassan12 Please set hparams as those in fastdiff.yaml
.
For pre_align_cls, use pre_align_cls: egs.datasets.audio.pre_align.PreAlign
.
For binarize, use binarizer_cls: data_gen.tts.vocoder_binarizer.VocoderBinarizer
I got error importing those files
Hi, I have updated the config file, please try again using modules/FastDiff/config/FastDiff.yaml
.