-
I've set up benchmarks for the reverse mode and forward mode autodiff for varying parameters.
My plan is to run benchmarks for both decents for Matrix sizes varying from 3 by 3 up to 61 by 61, with…
-
I am on Julia 1.8.5 and Enzyme 16818fd3d39d0583915aee38595b54a7fcce6b58. The following errors, though forward mode works.
```julia
using Enzyme, Statistics
f(x) = middle([2.0, x, 1.0])
autodiff(Re…
-
I tried to differentiate any of the following two functions using fwd and reverse mode.
The solver is marked as inactive and I try to get the gradients of the control variable.
```
void Callable(…
-
#### Summary:
Replace `var_alloc_stack_` data structure and rewrite functions that depend on it
* [ ] replace all the functions that depend on `var_alloc_stack_` with explicit allocations wrappe…
-
## Description
Add adjoint-Jacobian specialization for reverse mode for the fast Fourier transform (FFT) and its inverse.
## Example
#### FFT case
If `y = fft(x)`, then the adjoint-Jacobi…
-
MWE:
```julia
using DiffEqFlux, Zygote, DifferentialEquations, LinearAlgebra
k, α, β, γ = 1, 0.1, 0.2, 0.3
tspan = (0.0,10.0)
function dxdt_train(du,u,p,t)
du[1] = u[2]
du[2] = -k*u[1] …
-
Hi. I'm sorry if this question is due to a basic misunderstanding of how the AD in JAX works. I tried to work through the code but found it pretty tough to follow.
Basically, for vjps we get back a…
-
For some applications like object tracking it becomes tedious to write the jacobian to use within a Stan program. See https://discourse.mc-stan.org/t/jacobian-gradient-function/5614.
The function …
-
I've found that the gradients from `adjoint_sensitivities` are incorrect when the `g` function depends on the parameters `p`.
Consider using an `ODEProblem` to describe the state of a continuous ti…
-
Thanks for the awesome library! It would be great if PyTorch could support forward-mode automatic differentiation. The main use case is to compute a Jacobian-vector product. I tried using [this trick]…