Calling Flux.gpu on a view dumps core

[jessebett]

using MLDataUtils
using CuArrays
using Flux

x = rand(4,10)
x_b = batchview(x,7)
x_g = gpu.(x_b) # works

x_g = gpu(x_b) # aborts with core dumped

Not clear if this is a CuArrays problem or a Flux.gpu issue, but if you call gpu on a view, like provided by MLDataUtils.batchview broadcasted it will abort with (core dumped)

[maleadt]

Going to close this stale issue. If something like this still happens, please open a new issue with updated information