jessebett

`clip_by_value` is causing `train.minimize` operation to throw an error: `UndefRefError: access to undefined reference` ``` var = Variable(fill(0.1,T-1)) #

Rules for FFT

8

It appears there are adjoint rules for `FFT` in Zygote here: https://github.com/FluxML/Zygote.jl/pull/215

Insert pagebreak and inserting other html

3

I would like to insert a pagebreak when weaving to pdf. From https://stackoverflow.com/questions/22601053/pagebreak-in-markdown-while-creating-pdf#29642392 it appears that standard markdown has a way to insert raw html/css which would allow: `` However,...

Fix integer partitions

6

This would close #79 #80 and #81. Solution is due to @simonschoelly. Note that `integer_partitions` is still implemented, but now just `collect`s `partitions(n::Int)`. However, this implementation now returns the partitions...

partitions(0) does not return empty set

3

Currently ``` julia> partitions(0)|>collect 1-element Array{Any,1}: #undef ``` This is unexpected because the [partition of the empty set is the empty set](https://en.wikipedia.org/wiki/Partition_of_a_set#Examples) so I would expect ``` patitions(0)|>collect == [[]]...

`integer_partitions(n)` should be deprecated in favour of `partitions(n::Integer)`. Additionally, `partitions` is not documented in the readme. These functions currently do not behave the same: In the 0 case, both are...

`partitions(-1)` should produce an error when it is constructed: ``` # error here julia> partitions(-1) Combinatorics.IntegerPartitions(-1) julia> partitions(-1)|>collect ERROR: ArgumentError: destination has fewer elements than required Stacktrace: [1] copyto!(::Array{Any,1}, ::Combinatorics.IntegerPartitions)...

Student Roster Linking Process is Incredibbly Frustrating

5

Linking students from my roster to their github account is incredibly frustrating for students and me as the instructor. Not only does it expose the entire course roster (which is...

Fast Convolutions and Performance in NNlib

6

Convolutions provided by the [`FastConv` package](https://github.com/aamini/FastConv.jl) Described in [their paper](https://arxiv.org/pdf/1612.08825.pdf) is considerably outperforming the back ends for 1D and 2D convolutions. At least on CPU. ```julia using FastConv using NNlib...

@staticfloat we discussed some convenience functions for specifying `ConvDims` Here's a couple: Defining method for `spatial_dims(u::AbstractArray)` which gives the spatial dimensions for any array following the expected convention `(spatial_dims..., channels,...