-
@hzfan and I found `topi.nn.conv2d_transpose_nchw` and `te.gradient` behave strange when they are combined together and kernel's (height, width)=(1, 1).
For instance,
```
X = te.placeholder((8…
-
Hamiltonian Monte Carlo (HMC) is a widely used, gradient based, MCMC algorithm, that is the backbone of Stan's inference. I plan to implement it for monad-bayes. Todos (checkboxes indicate things done…
-
By "vector mode" I mean the ability to propagate a chunk of seeds in the pullback, similar to ForwardDiff's `Chunk` or Enzyme's `BatchDuplicated`.
Thanks to https://github.com/gdalle/DifferentiationI…
-
I understand the purpose of [`autodiff_thunk` in reverse mode](https://enzymead.github.io/Enzyme.jl/stable/api/#EnzymeCore.autodiff_thunk-Union{Tuple{RABI},%20Tuple{ModifiedBetweenT},%20Tuple{Width},%…
-
Found in https://github.com/SciML/LinearSolve.jl/pull/377, the Krylov methods have their own issue. MWE:
```julia
using Enzyme, ForwardDiff
using LinearSolve, LinearAlgebra, Test
n = 4
A = ra…
-
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
…
-
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…
-
I'm curious if your visions include making it a feature-complete NN training framework?
What will be the master plan? Integrating with Torch/TF/MXNet or build hardware-level compilation framework f…
-
Hi! Almost there. Most of the tests passed except for those involving `Float32`. Here is a MWE:
```julia
using Distributions
using DiffResults
using LinearAlgebra
using SimpleUnPack: @unpack
u…
-
I'm restarting the discussion on parameter servers from https://github.com/dask/dask-glm/issues/57
When doing distributed training it is often useful to rapidly update and share parameters between …