Error using CuIterator with Flux.train!

Specific Domains GPU
1

I’m trying to use CuIterator and Flux.train!. Attempting to do so gives an error because train! tries to get the length of the CuIterator. So this:

iter = CuIterator(batch(i) for i in 1:BATCH_COUNT)
Flux.train!(loss, ps, iter, opt)

results in an error:

MethodError: no method matching length(::CuIterator{Base.Generator{UnitRange{Int64}, var"#17#18"}})

It appears that HasLength() is the default for Base.IteratorSize and CuIterator doesn’t specify it, so it uses the default.

I can workaround it by setting it, but… should this be in CUDA.jl, or is there some other way I should be doing this?

 Base.IteratorSize(::Type{<:CuIterator}) = Base.SizeUnknown()
2

I have seen the same but forgot to make an issue/pr. I think the correct(er) fix is to delegate both HasLength and length to the wrapped iterator.

3

Right. Thanks for your help! I created an issue.