I have to ask that I have upgraded the DeepSpeech from 0.6.1 to 0.7.4.
The training check-points I have tried to use from the 0.6.1 last best trained/validated points.
I have seen that the values of loss is very high and computation is very slow. Can you comment why it could be? rest parameters I have used is all old.
Again, look at how to ask and post more info. I have no idea what your are doing.
lissyx
((slow to reply) [NOT PROVIDING SUPPORT])
3
You have been around on the forum for a while now, and you still don’t care enough to produce actionable informations so we can help you. This is wasting everyone’s time and energy. Please be responsible.
Dear @lissyx@othiele Thank you so much for kind help and be kind always.
I have just changed the DeepSpeech version. my training data is same, I restarted the training from the last saved checkpoints. I am investigating maybe my cuda broke and it is not getting proper load. Once the cuda is properly under control working, I will let you know if still problem exists. Thank you again.
lissyx
((slow to reply) [NOT PROVIDING SUPPORT])
5
Ok, take note: we have a lot of people asking for help on multiple matters, beside our own data / code to share.
So we can’t remember of each and everyone’s specifics.