Carlo Lucibello comments

Results 636 comments of


                                            Carlo Lucibello

Incorrect gradient of convolution w.r.t. weights

@dfdx for me the example script works fine on NNlib v0.6.6. Could you check if the problem still persist for you?

Incorrect gradient of convolution w.r.t. weights

> @CarloLucibello In your case, does the code work and show `true` in both 2 lines? yes, I get `true` in both lines. weird

Incorrect gradient of convolution w.r.t. weights

NNPACK looks like the only possible culprit. I don't have it ```julia julia> NNlib.is_nnpack_available() false ```

Swin Transformer related utility function

@XWHtorrentx it is still not clear what you are trying to accomplish with this PR. As @darsnack said, we don't want to provide utility functions in this repo, but just...

Use existing implementations in StatsFuns

why StatFuns's functions allow for array input?

Use existing implementations in StatsFuns

this shouldn't have any impact on zygote or gpu support, right?

Added the FastAI UNet implementation

A slightly modified UNet implementation can also be found in https://github.com/FluxML/model-zoo/pull/349

(Flaky?) CI failures on GHA latest + Buildkite

Does someone understand the buildkite error we are having? https://buildkite.com/julialang/nnlib-dot-jl/builds/306#2ff2a604-e66e-4ffc-99d4-abfc5e20be32/559-675

broken graph example on pyplot

@mkborregaard this works: ```julia julia> graphplot(A, node_weights = 1:n, line = (3, 0.5, :blues), marker_z = 1:n, marker=(:heat,:rect) ) ```

broken graph example on pyplot

I get this behviour on this setup ```julia # julia 0.5.1 julia> Pkg.status("PyPlot") - PyPlot 2.3.1 julia> Pkg.status("Plots") - Plots 0.10.3+ master julia> Pkg.status("PlotRecipes") - PlotRecipes 0.2.0+ master julia> import...