Hello, the DCGAN tutorial is broken in a few sense.
It is still using the old learning_parameter_schedule_per_sample method, which I cannot find any documentation anywhere on the web to replace it, so I simply replace it with learning_parameter_schedule in the learners package now
which actually brings us to the second problem
The result after 10000 mini batch is still purely noise, I can include the result if you need it.
I'm running CNTK 2.2, linux with a GTX1080, if it makes any different
Hello, the DCGAN tutorial is broken in a few sense.
I'm running CNTK 2.2, linux with a GTX1080, if it makes any different