Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign up
Highlights
- Pro
1,280 contributions in the last year
Contribution activity
June 2020
Created a pull request in google/jax that received 3 comments
- fix error when doing forward-mode of odeint
- removed stale faq entries
- update version and changelog for pypi
- allow closures for odeint dynamics functions
- add remat to top-level docs
- sketch out all-gather primitive
- avoid value-based error check in random.choice
- add back a full_lower, dropped in #3491
- fix an issue with newer versions of pytype
- remove the lower_fun default multiple_results=True
- pin numpy version in setup.py to avoid warnings
- ignore warnings in jax2tf tests
- fix jet typo: jnp not np
- refactor call primitives, simpler param processing
- use original numpy for `arange` shape calculations
- update tpu readme not to pin jax / jaxlib versions
- fix a bug from #3459
- simplify remat partial eval parameterization
- add systematic tests for vmap-of-pmap
- roll back of #3439 while we debug internal fails
- fix vmap-of-pmap mapped_invars logic bug
- add jax.numpy.concatenate(..., axis=None) support
- fix and better explain complex JVPs / VJPs
- print warning when doing jit-of-pmap
- add docstring to ravel_pytree
- Some pull requests not shown.
- jaxpr typechecking: custom rules and errors
- Fix lazy broadcast issue
- de-duplicate constants when lowering to XLA
- Remove unnecessary static_argnum in np.gradient
- Add jax.scipy.signal.detrend
- Cast int8 to bool for lax.not in jax2tf
- Fix typos: np.bool -> np.bool_
- Add class wrapper for double-double arithmetic
- doubledouble: Support repeated doublings
- implement jax.random.choice
- Add experimental precision doubling transform
- add jet primitives, refactor tests
- Jet rule for `integer_pow`
- Propagate raw __name__ of functions wrapped by jit and sharded_jit.
- jaxpr pretty-print: wrap equation RHS when the LHS is long
- add erf_inv rule
- Fix init step sizes to account for the time span of integration
- Document valid enum values for precision.
- Attach source info to Jaxpr equations.
- Initial version of invertible AD implementation
- allow scalar input in poisson sampler
- factor autodiff and vmap tests out from lax_test
- reuse input/output values when joining partially evaluated conditional branches
- Switch CI builds from Travis to Github actions
- Remove special case for scan masking
- Some pull request reviews not shown.
Created an issue in google/jax that received 1 comment
make pmap docs point more clearly to collectives
At the top level of jax.readthedocs.io, we have a section on pmap titled "Parallelism", but that doesn't include any obvious link to the documentat…
1
comment