Florian Metze
Great, thanks. Would you like to prepare a pull request to make sure we get thr correct fix?
Not sure - this all seems fine to me? Where is the error?
The logs show EPOCH 25 RUNNING - presumably you have models from epoch 0 (untrained) to 24 - not? Or are you for some reason starting the training at epoch...
I believe this means that you have an invalid entry in your dictionary or your vocabulary somewhere. Please check.
Did you look at the training logs of the 9th iteration. Do you have overflows, NaNs, or simialr events?
All your weights are completely saturated in each LSTM layer, i.e. you have min=-50 and max=50 in the networks. You can reduce this with --max-grad (or so) when you create...
Hi, my guess is that you will need to reduce the number of parameters in the model - l=5 and c=320 are good settings for Switchboard and TEDLIUM, with hundreds...
We have not run such experiments. I think there is some work on how to build uni-directional LSTMs that work for speech (mainly stacking future frames rather than relying on...
Not sure if this is an eesen-specific thing. Have you seen other code that shows this error? In any case, it would be helpful to have more information about what...
Siddharth, did you add documentation for this to the Eesen repository yet? Thanks, Florian > On Mar 8, 2018, at 12:58 AM, Minesh Mathew wrote: > > I understand that...