-
Will there be a `value_and_pushforward_split`?
-
```julia
using CUDA, Enzyme
w = cu(rand(Float32, 10, 10))
b = cu(rand(Float32, 10))
x = cu(rand(Float32, 10, 10))
act = x -> max(0f0, x)
function loss_function(act, w, x, b)
return sum(…
-
```
// #![feature(panic_unwind)]
#![feature(bench_black_box)]
// extern crate panic_unwind;
use autodiff::autodiff;
#[derive(Debug, Clone, PartialEq)]
struct Foo {
x: f64,
y: f64,
…
ghost updated
6 months ago
-
In a few PRs the level of testing and and review has been set sort of heuristically. I think it would be nice to have some levels established in a wiki somewhere for what level of testing and review i…
-
Reported in #7092. Related to the comment here: https://github.com/google/jax/blob/97a5719fcb40af7231b5f803f965063538282f8e/jax/interpreters/ad.py#L198-L200
The issue is that in the backward pass o…
-
I noticed the following change in https://github.com/gdalle/DifferentiationInterface.jl/pull/196:
Enzyme v0.11: type-stable
```julia
(@v1.10) pkg> activate --temp
Activating new project at `…
-
I couldn't find a better title, so let me explain: I have a function interpolated via BSplineKit, based on some lookup table. BSplineKit provides rrules for differentiation of the interpolated functio…
-
Click to expand!
### Issue Type
Bug
### Source
binary
### Tensorflow Version
tf 2.9
### Custom Code
Yes
### OS Platform and Distribution
Linux Ubuntu 20.04
### Mobile device
_No respons…
-
MWE:
```
using Distributions
using Enzyme
f(x) = sum(logpdf.((Skellam(x[1], 2.0),), -2))
Enzyme.gradient(Enzyme.Forward, f, [1.0])
```
Output:
```
[74169] signal (4.1): Illegal instructio…
-
Reverse-mode backends like [Zygote](https://fluxml.ai/Zygote.jl/dev/adjoints/#Pullbacks-1) and [Enzyme (?)](https://enzymead.github.io/Enzyme.jl/stable/api/#EnzymeCore.autodiff_deferred_thunk-Union{Tu…