Yoosif Sherif
Yoosif Sherif
Thanks man
I think there is a dataset called NEMLAR that is bigger than this one but has lots of problems
There are no plans to change. Having a good GPU and using it to synthesize the sound instead of the CPU would improve performance. I do not remember how much...
I am not an expert in the area, but I suggest that you google keywords like "best GPU deep learning"
@ishandutta2007 Aravec: https://github.com/bakrianoo/aravec Go to N-Grams Models and Unigrams Models sections For FastText, google FastText pre-trained model
@ishandutta2007 What are sample evaluate scripts?
It automatically gets created as the NN is being trained
@Shikherneo2 you mean sample_rate?
I think you could not use a checkpoint when hparams are changed.
Did you achieve alignment?