jcgeo9

@sooftware can you please answer to @zwan074 ? many of us are confused as to how to use a loss function to train the conformer as the outputs are log...

I have another question about the function of the conformer. I am using a vocab of 6030 classes and my input data: batch, dim, seq_len= 32, 201, 1162 (where 1162...

@sooftware When I execute the following code, recognize_sp variable has the shape: [32, 289] ``` import torch import time import torch.nn as nn from conformer import Conformer cuda = torch.cuda.is_available()...

@sooftware hmm ok but what do i do with that? i mean how do i convert it to what i actually want? isnt it suppose to return [32, 20] tensor...