Peter

Seems to be fixed already. Tested with Zygote 0.6.34: ```julia julia> Zygote.gradient(randn(3,3)) do x sum(sin.(cat(x; dims=4))) end[1] 3×3 Matrix{Float64}: 0.598398 0.988603 0.999602 0.84835 -0.101217 0.391286 0.9785 -0.717415 0.87662 ```

there is a [neon tutorial](https://github.com/NervanaSystems/neon_course) maybe you can take a look at this [one](https://github.com/NervanaSystems/neon_course/blob/master/04b%20Writing%20a%20custom%20layer.ipynb)

@guoxuesong in `bprop`, `self.deltas` is the error need to be back prop to the previous layer, alpha & beta is just parameters of `self.be.compound_dot`, you may want to take a...

It seems to be a problem when scalar appears and the broadcasted is too nested: ```julia using RecursiveArrayTools u = VectorOfArray([zeros(100, 100, 100), zeros(100, 100, 100)]); uprev = VectorOfArray([zeros(100, 100,...

@PallHaraldsson Hi, I don't really use TensorFlow.jl, but I need to read a bunch of Tensorflow checkpoint files in Julia and turn them in to BSONs (for my package Transformers.jl)....

Any update? Looks like the CI failed because it cannot find the shared library.

Like this: ```julia # sample 1 emb1 = randn(10, 5); idx1 = [1,2,3,3,4]; # sample 2 emb2 = randn(10, 7); idx2 = [1,2,2,3,3,3,4]; # batched sample emb = cat(hcat(emb1, zeros(10,...

> The only concern could probably be that while the GPU path is parallelised (it uses the same CUBLAS functions underneath as NNlib), the CPU path is not (https://github.com/FluxML/Metalhead.jl/pull/135#discussion_r829147436). I...

Which code are you testing with?

Generally, It might be better to wait for #103. > Which gives me following error: Model Val{:model_task}() doesn't support this kind of task: Val{:model_task}(). I tried different model tasks. I...