Steven G. Johnson

Couldn’t waitall be defined as a wait method that takes an array of requests, maybe with a flag argument for waitall vs waitany?

From the readme of `bipolor-colormap`: > This is not a modern well-designed colormap; it's not perceptually uniform, does not have uniform lightness steps, and the endpoints are not equal lightness....

It seems reasonable to support this via iteration order of `x`, i.e. column-major order. After all, that's the ordering that is used if `a` is multidimensional: ```jl julia> sample([17 3;...

(It won't give correct results for `replace=true` until #675 is fixed, however.)

Note that it is also easy to parse Jupyter notebooks and output `.jl` files yourself, because they are just JSON. See [this discourse post](https://discourse.julialang.org/t/export-jupyter-to-jl-does-not-preserve-markdown-cells/57556/2?u=stevengj) for example — replace `text/plain` with...

Also, libtensorflow_framework.so is installed by Conda… not sure why you need to download it separately? For the GPU?

Yes, 3rd arg adds the channel.

Exactly. Once you can define an operator that calls back to an arbitrary pure-Julia function, then you potentially get a whole bunch of things (like fusing broadcasts) for free. Note...

A complication is that `REGISTER_OP` in TensorFlow requires you to specify the input and output types, which doesn't map well onto Julia functions (that may allow multiple types). However, you...

Regarding gradients, with custom operations my understanding is that TensorFlow allows you to supply a gradient function too. e.g. you could automate this with ForwardDiff applied to the broadcast operand....