-
Luckily, adjoint methods for (most) eigenvalue problems tend to be rather trivial _if you are optimizing the eigenvalue directly_. So, for example, if you want to match the effective index of two mode…
-
The code refactoring in #92, #95, and #96 added much more robust and extensive fitting capabilities (even after removing the dependence on the very slow `astropy.modeling` classes) at the expense of s…
-
On v0.1.25 on OSX, I get the following error when computing gradients from the following jit-compiled function.
```python
import numpy as onp
import jax.numpy as np
from jax import grad, jit
…
-
The current implementation of Runge-Kutta with adjoint reverse-mode gradients is great, but there are a few things I still find myself missing, and I'd really love to help contribute, or just see in J…
-
The recently proposed [NEP-47](https://numpy.org/neps/nep-0047-array-api-standard.html) attempts to unify the APIs of various tensor frameworks (NumPy, Tensorflow, PyTorch, Dask, JAX, CuPy, MXNet, etc…
-
### Expected behavior
I was trying to compute the Hessian and saw that the Jax interface breaks down if we have the JIT on. Without JIT, it works fine. The error seems to be due to the non-availabi…
-
Was trying out the newly added Hessian-vector product function `hvp!` on a simple quadratic and I have run into the following strange error:
```julia
julia> using Enzyme: hvp!
julia> n = 10
…
-
Hi JAX developers,
I'm wondering if it's possible to calculate `jvp` for intermediate variables defined in a function. Here's an example:
```
def f(x, params_1, params_2, params_3):
W1, …
-
Consider the following program that tries to find values of x1 such that it is close to zero. After running one round of map_optimize, we should find x1 moved closer to zero (as in the subsequent trac…
-
This might be a documentation request, or a feature request.
Currently the AutoMALA constructor takes the argument `default_autodiff_backend`, which it uses via LogDensityProblemsAD to differentiat…