Conversation
| @test_broken collect(A * y) ≈ collect(A) * collect(y) | ||
|
|
||
| @test_broken gradient(A -> sum(abs, A * y), A)[1] isa CuArray # gather!(dst::JLArray, ...) fails |
There was a problem hiding this comment.
Is the solution here to use gather (and take on a dep)?
There was a problem hiding this comment.
I think this needs gather/scatter from NNlibCUDA to work on the GPU. And since there's no corresponding code for non-CuArray GPUArrays, I think it can't work with this fake JLArray.
For testing it, you could set up the whole buildkite story to run honest CUDA tests. But perhaps it's not worth it, and this package should just trust NNlib + NNlibCUDA to test things. And perhaps Flux to test the integration?
There was a problem hiding this comment.
Xref FluxML/NNlib.jl#427 too --- it would be nice if forgetting to load NNlibCUDA gave friendly errors, not scalar indexing.
It would be nicer if that could be loaded automatically, of course.
| if any(iszero, indices) | ||
| for x in data | ||
| isnothing(_findval(x, labels)) && error("Value $x not found in labels") | ||
| isnothing(_findval(x, labels)) && throw(ArgumentError("Value x = $x not found in labels = $labels")) |
There was a problem hiding this comment.
I changed these error types partly so that tests can distinguish scalar indexing errors from helpful messages.
| end | ||
| return OneHotArray(indices, length(labels)) | ||
| end | ||
| function _onehotbatch(data::AbstractArray, labels) # this works for GPUArrays too |
There was a problem hiding this comment.
One reason to change this is to avoid ever making a MVector or something weird like that:
| function _onehotbatch(data::AbstractArray, labels) # this works for GPUArrays too | |
| function _onehotbatch(data::AbstractGPUArray, labels) |
| indices = UInt32[something(_findval(x, labels), default_index) for x in data] | ||
| return OneHotArray(indices, length(labels)) | ||
| end | ||
| function _onehotbatch(data::AbstractArray, labels, default) |
There was a problem hiding this comment.
| function _onehotbatch(data::AbstractArray, labels, default) | |
| function _onehotbatch(data::AbstractGPUArray, labels, default) |
2 participants
Unlike FluxML/Flux.jl#1959, this uses
mapover arrays. Some duplication, unfortunately. Possibly the new method should be restricted to AbstractGPUArrays?Closes #16
Also tries to organise the tests just a little bit better.