-
MWE:
```julia
module MWE
import Bijectors, Enzyme, StableRNGs
b = Bijectors.PlanarLayer(3)
binv = Bijectors.inverse(b)
x = randn(StableRNGs.StableRNG(23), (3, 3))
f = x -> sum(b(binv(x)))
Enzyme.gradi…
-
When I use autodiff, an error message appears:
[verify.cpp:taichi::lang::IRVerifier::basic_verify@46] IR broken: stmt 8233 cannot have operand 8226.
when I close advanced_optimization, the error b…
-
The other day @janosg suggested that implementing some of the core algorithms in [JAX ](https://jax.readthedocs.io/en/latest/)might provide exceptional performance improvements.
`pyfixest's` alter…
-
We should add backends for the following AD/FD packages:
- [x] ForwardDiff
- [x] ReverseDiff
- [x] FiniteDifferences
- [ ] all ChainRules-supporting ADs (see #11, #39)
- [ ] FiniteDiff
- [x] Tra…
-
## Description
There are lots of functions that don't have reverse mode specializations.
#### Matrix operations
All of these would be very useful, espcially the basic arithmetic of the firs…
-
On Enzyme main (3c0871d54b85e72a8b0dc2e2e5d48d2eb6e0a95e) and Julia 1.10.3:
```julia
using Enzyme
f(x) = sum(diff([0.0 x; 1.0 2x]; dims=2))
autodiff(Forward, f, Duplicated(2.0, 1.0))
```
```
ERROR: St…
-
@llfung talked to me today about his work on adjoint-accelerated programmable inference for large PDEs, and what would be needed on Turing's part to support that. As I understand it (and I know very l…
-
Hi, is there a way one could specify `nondiff_argnums` for the custom vjp of a primitive? Like using `jax.interpreters.ad.defvjp_all` or `jax.interpreters.ad.defvjp`?
-
### 🐛 Describe the bug
```
Traceback (most recent call last):
File "main.py", line 202, in
all_metrics = trainer.train(args.steps, args.val_steps, args.save_every, args.eval_every)
Fil…
-
I diagnosed this problem with DifferentiationInterface, but it also appears using Enzyme's own `hvp` function.
The following code is run on Julia 1.11 with Enzyme v0.13.14. On Julia 1.10 it returns th…