1051: add DataLoader r=CarloLucibello a=CarloLucibello
Fix#450
This adds a DataLoader type, largely adapted from the Knet one, therefore pinging @denizyuret to check if he is cool with this. Unfortunately, I cannot "unsee" his implementation, and in any case any reasonable alternative implementation will be pretty much similar I guess.
This is an initial implementation to get things going, possibly in the future we will also want a distributed and out-of-memory option as the one implemented by @staticfloat here
https://github.com/FluxML/Metalhead.jl/blob/sf/training/training/ImageNet/dataset.jl
Co-authored-by: CarloLucibello <carlo.lucibello@gmail.com>
1023: Feature: Added Boston Housing Dataset r=CarloLucibello a=pranjaldatta
[Boston Housing Dataset](https://archive.ics.uci.edu/ml/machine-learning-databases/housing/) is one of the most common datasets that are used by beginners. It is as popular as other datasets like Iris etc. Hence, it feels only natural that this dataset is a part of Flux.
Added src/data/housing.jl: code for downloading and loading the dataset
Edited src/data/Data.jl: To include and export housing.jl
Edited test/data.jl: Added test for the dataset.
*All tests in test/data.jl are passing*
Co-authored-by: pranjaldatta <pranjaldatta99@gmail.com>
Co-authored-by: Pranjal Datta <pranjaldatta99@gmail.com>
1041: add NNlib docs + misc docs improvements r=CarloLucibello a=CarloLucibello
Partially addressing https://github.com/FluxML/NNlib.jl/issues/137.
Also, I'm leaving out the `σ` activation and using its alias `sigmoid`, since `σ` conveys little information and it is also used to denote a generic activation in the Dense layer. I think we should deprecate `σ` in NNlib, has this been discussed already?
In an ideal world, before merging this, we should get NNlib to either unexport or add docs to its undocumented exports
Co-authored-by: Carlo Lucibello <carlo.lucibello@gmail.com>
998: test restructure on the GPU r=CarloLucibello a=ChrisRackauckas
Requires https://github.com/FluxML/Zygote.jl/pull/474 to pass
Co-authored-by: Chris Rackauckas <accounts@chrisrackauckas.com>
960: Added utility function outdims to compute output dimensions of a layer r=dhairyagandhi96 a=darsnack
Based on Slack chatter, I added a utility function, `outdims`, that computes the output dimensions for given input dimensions.
Example
```julia
layer = Conv((3, 3), 3 => 16)
outdims(layer, (10, 10)) # returns (8, 8)
```
Co-authored-by: Kyle Daruwalla <daruwalla@wisc.edu>
1013: Adapt to GPUArrays/CuArrays changes r=dhairyagandhi96 a=maleadt
Changes in response to a29df67184 and https://github.com/JuliaGPU/CuArrays.jl/pull/576. I suppose the next CuArrays release will need to be breaking because of this.
Maybe the `crossentropy` signature needs to be adjusted to support integer vectors, but I'll leave that decision up to Flux developers. This at least is the quick fix to get the tests passing again.
Co-authored-by: Tim Besard <tim.besard@gmail.com>