Kai Wang
> Question 1: > In following codes: > train_clean_path = '/media/concordia/DATA/KaiWang/pytorch_learn/pytorch_for_speech/dataset/voice_bank/trainset/clean_trainset' > train_noisy_path = '/media/concordia/DATA/KaiWang/pytorch_learn/pytorch_for_speech/dataset/voice_bank/trainset/noisy_trainset' > train_mix_path = '/media/concordia/DATA/KaiWang/pytorch_learn/pytorch_for_speech/dataset/voice_bank_mix/trainset' > > I know that train_clean_path‘s input is clean trainset,but i...
> I want to know what is the inputs of test_log_path ? test_log_path= '/media/concordia/DATA/KaiWang/pytorch_learn/pytorch_for_speech/dataset/voice_bank/log/logfiles/log_testset.txt' Looking forward to your reply.Thank you very much! Hi, thank you very much for your questions....
Hi, thank you for reaching out. In my implementation, I set batch size as 2 and each epoch is iterated about 5100 steps. The training time of each epoch is...
> I also use GTX 1080ti gpus. The dataset is the noisy dataset of the DNS Challenge I synthesized, not the Voice_Bank dataset with 28 speakers.And I haven’t changed your...
> 请问您这套模型的实时性怎么样呢？ 您好，很抱歉现在才回复您。我在做这套模型的时候还没有测实时性。您如果要测的话，可以把attention 换成masked attention去屏蔽future information， 同时把global transformer中GRU变成单向，希望可以帮助到您。
Hi, please use the updated dataset link.
> clean from VCTK：https://datashare.ed.ac.uk/handle/10283/2950 ?? noise form DEMAND：https://zenodo.org/record/1227121 ?？ Hi, thank you very much for reaching out. Please use this dataset link: https://datashare.ed.ac.uk/handle/10283/2791 This dataset has included the clean and...
Hi, sorry for late reply. If you want to test the real-time or casual performance of TSTNN, please try to do following modifications: 1. Using masked attention in transformer for...
> Thank you very much for your questions. I will update the details in Readme.txt since I am very busy now. Thank you very much for your questions. I will...
> Hi, thank you very much for your questions. Please use the VCTK train_28spks to train the TSTNN. Because current version of TSTNN is low footprint. If you want to...