Motivation The appeal of automatic differentiation is that you can write a function using ordinary Julia and the gradient is automatically computed. Disallowing mutation in that function seems reasonable on the surface—it seems like it should be pretty easy to write code that doesn’t mutate arrays. …

Idea to make Zygote support mutation in easy cases

Tomas_Pevny December 26, 2022, 5:12am 2

Have you checked Zygote’s buffers? It seems they do exactly what you suggest.

Topic		Replies	Views
Zygote.gradient(): Mutating arrays is not supported General Usage	1	794	August 18, 2020
Gradient of a loss function : struggling to avoid arrays mutation New to Julia zygote , sciml	4	1505	December 7, 2020
Zygote debugging General Usage	6	2107	October 1, 2019
Gradient of gradient Machine Learning	9	1299	November 6, 2020
Mutating versus non-mutating arrays for Zygote Gradient General Usage	6	383	December 25, 2022