I can't launch the training on long audio files. I had the following error "sum of alphas * betas (5.034e-24 <> ....."
So I change the value of S2_ALPHA_BETA_EPSILON from 0.01 to 0.1
Now, I can launch the training on long audio files. Does the variable change something to the recognition quality ?
Log in to post a comment.
I can't launch the training on long audio files. I had the following error "sum of alphas * betas (5.034e-24 <> ....."
So I change the value of S2_ALPHA_BETA_EPSILON from 0.01 to 0.1
Now, I can launch the training on long audio files.
Does the variable change something to the recognition quality ?