my training parameters
number of hidden units 1024 (tried with 256 as well)
batch size 110 (tried with lower batch size as well)
loss saturates at 27 for few epochs and then starts to increase again
360 hours of training data
my training parameters
number of hidden units 1024 (tried with 256 as well)
batch size 110 (tried with lower batch size as well)
loss saturates at 27 for few epochs and then starts to increase again
360 hours of training data