AutoGrad.jl Versions Save

AutoGrad v1.2.4

Diff since v1.2.3

Closed issues:

• Alkan bias add problem? (#121)

AutoGrad v1.2.3

Diff since v1.2.2

Closed issues:

• getindex gradient collection bug (#116)

AutoGrad v1.2.2

Diff since v1.2.1

Merged pull requests:

• Install TagBot as a GitHub Action (#118) (@JuliaTagBot)

v1.2.1 (2020-02-01)

Diff since v1.2.0

v1.2.0 (2019-10-25)

Diff since v1.1.6

Closed issues:

• Autograd.Sparse type causes regression (#114)
• Extend UngetIndex to multiple indices, values for efficiency (#112)

Merged pull requests:

• Fixing #114 Sparse bugs (#115) (denizyuret)

v1.1.6 (2019-09-30)

Diff since v1.1.5

Merged pull requests:

• Sparse gradients (#113) (denizyuret)

v1.1.5 (2019-09-20)

Diff since v1.1.4

v1.1.4 (2019-05-25)

Diff since v1.1.3

Merged pull requests:

• Dy/highorder (#111) (denizyuret)

• Use broadcasted rather than broadcast as primitive, allows p .-= g where p::Param.
• Define efficient params(::Tape).
• doc updates.

• Support for broadcasting user defined functions.
• gcheck and @gcheck for gradient checking with Params.
• Added @primitive2 and @zerograd2 for broadcast-only primitives.
• Handle functions where result is not the last thing on tape.