Highlights
- 37 discussions answered
Pinned
1,221 contributions in the last year
Less
More
Contribution activity
May 2021
Created 9 commits in 2 repositories
Created a pull request in google/jax that received 3 comments
Opened 9 other pull requests in 1 repository
google/jax
6
merged
3
open
- CI: test Python 3.9 with numpy dispatch test
- [sparse] add batching rules to COO primitives
- Sparse AD: add transpose rules for COO primitives
- Improve error message for shape
- DOC: add documentation of DeviceArray object properties & methods
- [sparse]: add multi-buffer implementation
- Cleanup: remove unnecessary utility
- WIP: lower unit to nothing
- DOC: add bicgstab to doc page
Reviewed 18 pull requests in 1 repository
google/jax 18 pull requests
- Initial implementation of jax.numpy.poly
- JVPs for sparse ops.
- Support complex numbers in jnp.convolve and jnp.correlate.
- Update documentation for custom_jvp handling of nondiff_argnums as ar…
- Add regression test for #5728.
-
Reduce boilerplate of
FrozenDict - Add version constraints to flatbuffers versions.
- Reenable numpy dispatch tests.
- Use Array instead of jnp.array in jax.random
- Promote the x.at[idx].set(y) operators as the preferred way to do ind…
- Allow non-inexact dtypes for jnp.nan...() reductions.
- Improve type error when an object dtype is passed to an operator like +.
- Update code example in custom VJP documentation.
- Make jnp.unravel_index rank promotion explicit.
- Do not perform second call to config_with_absl
- Add associated Legendre functions of the first kind
-
Support
dtypekwarg injnp.matmul - Fix pip install instruction for forks in contributing section

