Skip to content

Releases: google/neural-tangents

v0.6.5

11 Dec 14:10
Compare
Choose a tag to compare

v0.6.4

25 Aug 00:01
Compare
Choose a tag to compare

v0.6.2

v0.6.1

14 Sep 04:58
Compare
Choose a tag to compare

v0.6.0

18 Jul 19:09
Compare
Choose a tag to compare

v0.5.0

23 Feb 16:58
Compare
Choose a tag to compare

Potentially breaking changes:

New features:

v0.4.0

17 Feb 00:57
Compare
Choose a tag to compare

WARNING:
Our next major release (v0.5.0) will include significant refactoring, and could break your code if you use internal function like nt.utils.typing, nt.utils.utils, nt.utils.Kernel etc. (public API will remain unchanged). This should be easily fixed by updating the imports, e.g. nt.utils -> nt._src.utils.

This release (v0.4.0):

New feature:

Improvements:

Bugfixes:

Breaking changes:

  • Bump requirements to jax==0.3 and frozendict==2.3.

v0.3.9

v0.3.8

07 Oct 00:14
Compare
Choose a tag to compare

New Features:

  • stax.Elementwise - a layer for generic elementwise functions requiring the user to specify only scalar-valued nngp_fn : (cov12, var1, var2) |-> E[fn(x_1) * fn(x_2)]. The NTK computation (thanks to @SiuMath) and vectorization over the underlying Kernel happen automatically under the hood. If you can't derive the nngp_fn for your function, use stax.ElementwiseNumerical. See docs for more details.

Bugfixes:

Full Changelog: v0.3.7...v0.3.8

v0.3.7