[theano-users] Theano dtensor3 type variable gives error in GPU

[shyam krishna khadka]

I have code that is running smoothly in cpu with a variable defined as :

l = T.dtensor3('l')





Please help.

[shyam krishna khadka]

I have code that is running smoothly in cpu with a variable defined as :

l = T.dtensor3('l')


But when I ran it in NVIDIA GPU, it gives error as :

TypeError: Cannot convert Type TensorType(float32, 3D) (of Variable Subtensor{int64:int64:, ::}.0) into Type TensorType(float64, 3D). You can try to manually convert Subtensor{int64:int64:, ::}.0 into a TensorType(float64, 3D).

It seems the datatype is float64. But for GPU, we need to make it float32. What will be equivalent for this in GPU ? Itired with

l = T.tensor3('l')

But now I am getting assertion error as

A.ndim == 3 error