Skip to content

Releases: mashu/PositionalEmbeddings.jl

v0.6.2

05 Dec 09:15
Compare
Choose a tag to compare

PositionalEmbeddings v0.6.2

Diff since v0.6.1

Exclude LinearAlgebra compat and relax CUDA compat.

v0.6.1

04 Dec 20:41
Compare
Choose a tag to compare

PositionalEmbeddings v0.6.1

Diff since v0.6.0

Relax LinearAlgebra default compat that was set automatically.

v0.6.0

04 Dec 13:26
Compare
Choose a tag to compare

Full Changelog: v0.5.0...v0.6.0

v0.5.0

26 Nov 15:06
Compare
Choose a tag to compare

Full Changelog: v0.4.0...v0.5.0

v0.4.0

26 Nov 12:05
Compare
Choose a tag to compare

PositionalEmbeddings v0.4.0

Diff since v0.2.0

Make compat less strict.

v0.3.0

26 Nov 12:04
Compare
Choose a tag to compare

PositionalEmbeddings v0.3.0

Update to match reference implementation from https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/transformers/rope/

Merged pull requests:

  • build(deps): bump codecov/codecov-action from 4 to 5 (#1) (@dependabot[bot])

Closed issues:

  • TagBot trigger issue (#2)

v0.2.0

26 Nov 12:04
Compare
Choose a tag to compare

PositionalEmbeddings v0.2.0

Fix RoPE neg_half to correctly work with pairs.

Merged pull requests:

  • build(deps): bump codecov/codecov-action from 4 to 5 (#1) (@dependabot[bot])