This print the time used for each batch.

The old range was causing nan in float32.

When the output was of size 0, we where returning a numpy ndarray
instead of a CudaNdarray.

No news, as this is a fix to something new since the last release.

Make GpuCrossentropySoftmaxArgmax1HotWithBias and GpuCrossentropySoftmax1HotWithBiasDx work for bigger inputs.

I don't know if argsort should have a grad defined, so I mark it as not implemented.

Fix opt crash of local_gpu_lazy_ifelse.

Also print the inputs type.

exception_verbosity now only add the debugprint that now also have the print_type parameter

Fix GpuConv error reporting. The error was raised in the next executed apply node when we detect the input to be None instead of CudaNdarray.

this fixes gh-786.

It was passing event if it tested the wrong stuff. After fixing it
still work correctly.

It was a set, not it is now a list. This cause problem in the error handling code where we try to get the index in a set. That isn't defined.

Reported by Ryan Price

Gpu iadd for 6d tensor

Gpu extract diag

Better error message when an apply node execution fail.

fix gh-1133, don't replace error message from compute_test_value.

Fix sum sum crash

fix garray_to_cudandarray for dimensions of 0. CudaNdarray need a strides of 0 in those cases, not 1. Error was raised in filter about this.