-
We could preload them using `rirud` only in C++ layer but not initialize them to Java environment.
This will slightly decrease disk reading and decrease memory usage.
-
I am just copying from https://github.com/FluxML/Zygote.jl/issues/668. Not sure if this should be fixed here or in Zygote.
```julia
julia> using Zygote, ProgressMeter
julia> function train()
…
-
EDIT: PLS, check my update on the reply below.
This is what I get in stock oos, rooted with magisk, and stock kernel on the oneplus 5, xposed v88.1, but it was the same in v88.
I get this even w…
ghost updated
5 years ago
-
Looks like the time-to-solution to obtain gradients from TEBD simulation of dynamics (i.e. OCC) have substantial overhead compare to simple evaluation of the function.
-
MWE:
```julia
julia> using LinearAlgebra, Zygote, CUDA
julia> Zygote.gradient(x -> tr(x), CUDA.zeros(2,2))[1] # works
2×2 Diagonal{Float32, FillArrays.Fill{Float32, 1, Tuple{Base.OneTo{Int64}}…
-
For several utilities, benchmarking suggests that different backends given different gradients. This should be investigated as it may indicate performance issues.
-
Hello,
A collegue pointed me to your package and it seems like it might be a useful method for speeding up parameter estimation for some of our models. Thank you for putting this together.
I h…
-
Currently `Zygote.gradient` projects the co-tangent returned from `pullback` to have the same sparsity structure as the input. This is mathematically incorrect when the matrix input is sparse. Accordi…
-
### Motivation and description
When differentiating something complicated which contains mutation, it can be hard to know exactly where the mutation is. In this example, the mutation is tucked away…
-
I want to compute the gradient of a loss function with respect to a `ComponentArray`. In the loss function, I need to reconstruct a `ComponentArray`.
Based on @jonniedie reply https://github.com/jon…