I am training my acoustic model which is multilayer DNN written in PyTorch. I want to save the output of the network as (.ark) file to use it in the decoding stage for automatic speech recognition. I searched a lot but can't find a function to do this. Any Help !!!
Asked
Active
Viewed 82 times
0
-
ark is format for features and other data files, not for neural network models. You probably want to do something else – Nikolay Shmyrev Apr 19 '20 at 12:02
-
I want to save the network output in a stdout file so is there is a function for this or can you tell me if I have a tensor how can I save this tensor in .ark file format – Mohamed Nabih Apr 19 '20 at 13:07
-
kaldi-io python package should be able to do that – Nikolay Shmyrev Apr 19 '20 at 14:06