Yota.jl is a package for reverse-mode automatic differentiation in Julia. The main features are:
- optimized for large inputs and conventional deep learning
- tracer-based with a hackable computational graph (tape)
- supports ChainRules API
- Popularity
- 158 Stars
- Updated Last
- 7 Months Ago
- Started In
- July 2018
Required Packages
-
Adapt
-
Atomix
-
CEnum
-
ChainRules
-
ChainRulesCore
-
Compat
-
ConstructionBase
-
DataAPI
-
DataValueInterfaces
-
Distributed
-
ExprTools
-
FiniteDifferences
-
GPUArrays
-
IrrationalConstants
-
IteratorInterfaceExtensions
-
KernelAbstractions
-
LLVM
-
NNlib
-
OrderedCollections
-
PrecompileTools
-
Preferences
-
RealDot
-
Reexport
-
Requires
-
Richardson
-
SparseArrays
-
SparseInverseSubset
-
StaticArrays
-
StaticArraysCore
-
StructArrays
-
Tables
-
TableTraits
-
Umlaut
-
UnsafeAtomics