-
```
m = mortar([rand(i,j) for i=1:3, j=2:3])
m .- (1:5)'
```
In 0.16.22-23:
```
ERROR: BoundsError: attempt to access 1-element Vector{Int64} at index [2]
Stacktrace:
[1] getindex
@ .\a…
-
Currently weighting is in a somewhat pitiful state in ODL. We support it in principle, but basically none of the operators and functionals get it right.
Example:
```python
>>> rn = odl.rn(3)
>>>…
-
The current tolerances are huge:
https://github.com/AMICI-dev/AMICI/blob/bea2dfe37040e7a6f36f163a46e03e428bab75cf/tests/benchmark-models/test_petab_benchmark.py#L14-L16
-
I was surprised that `Diagonal` turns `1×n` adjoints into `1×1` diagonals:
```julia
julia> Diagonal(rand(2)')
1×1 Diagonal{Float64,Array{Float64,1}}:
0.05075540276803281
```
That's because …
-
Some things that probably can be improved as a follow-up to the initial implementation (but might require changes in other packages such as AD backends):
- Avoid dense matrices (currently triangular …
-
As the title says. Code to see this:
```julia
using CUDA, Zygote
function test_func(a, b)
return sum(abs2, a .+ b')
end
a = CUDA.rand(ComplexF64, 3)
b = CUDA.rand(3)
gradient(test_func, a,…
-
- S-duality in 4d N=2
- Seiberg duality in 4d N=1, generalizations by Kutasov-Schwimmer (one adjoint), Brodie (two adjoints), Intriligator-Pouliot (symplectic), Argyres-Intriligator-Leigh-Strassler (t…
-
FMI 3.0 can now be generated with tools such as Dymola. Importantly, FMI 3 introduces support for adjoint derivatives, which makes embedding FMUs in CasADi expressions potentially much more efficient.…
-
- [x] Optimise-then-discretise
- [ ] Reversible solvers
- [ ] Interpolating adjoints
- [x] Checkpointing?
- [ ] Quadrature?
-
### Feature details
PennyLane ships a few useful transform, but some obvious one that would be nice to have are missing. These include for example transforms that:
* Combine all all GlobalPhase ga…
cvjjm updated
1 month ago