jakevdp
Follow
jakevdp
Follow
Highlights
- 176 discussions answered
Pinned
1,546 contributions in the last year
Contribution activity
June 2022
Created 16 commits in 1 repository
Created a pull request in google/jax that received 3 comments
[x64] deprecate unsafe type casting in scatter-update operations
Fixes #10892 This is a breaking change – at the very least, we'll need to fix some internal tests and some downstream users. Also, perhaps this sho…
+107
−16
•
3
comments
Opened 19 other pull requests in 1 repository
google/jax
3
open
16
merged
- [x64] make lax_scipy_test.py compatible with strict dtype promotion
- WIP: [x64] preserve weak types in promote_dtypes_inexact
- [x64] gmres: avoid problematic type promotions
- [sparse][x64] prevent unnecessary dtype promotion in sparse impls
- DOC: add FAQ section on zero gradients for rank-based operations
- lax_numpy_test: test bitwise ops on full input range
- [x64] Make TPU svd compatible with strict type promotion
- jnp.unique: improve error when run under JIT
- CI: unpin traitlets
- lax_numpy_test: test bitwise ops on full input range
- [x64] make jnp.modf() compatible with strict dtype promotion
- [x64] make jax.numpy reductions respect input dtypes
- [x64] make jnp.insert compatible with strict dtype promotion
- [x64] make jnp.corrcoef compatible with strict dtype promotion
- [x64] make jnp.average compatible with strict promotion
- [x64] make linspace, logspace, & geomspace compatible with strict promotion
- [x64] make jnp.interp safe for use with strict dtype promotion
- [x64] make jnp.mgrid compatible with strict dtype promotion
- [x64] jnp.ldexp: avoid implicit 64-bit promotion
Reviewed 13 pull requests in 1 repository
google/jax
13 pull requests
-
Add bitwise XOR reducer to
lax.reduce - Extend Common-Gotchas docs: speed of jnp.array
- Added scipy.stats.gennorm.
- DOC: add FAQ section on zero gradients for rank-based operations
- Annotate vmap
- Add jax.default_device to CHANGELOG
- Update testReducerNoDtype and others to not run duplicate tests.
- Added random.gnormal and random.ball.
- WIP: Implementation of Scipy Bootstrap
-
add
scipy.special.factorialandscipy.special.comb - Disallow complex dtypes for lax ops rem, lt, le, gt, and ge.
- Use util.unzipN() in more places instead of zip(*args).
- Fix autodiff of pad() of scalar inputs.
Created an issue in google/jax that received 4 comments
Abort trap in jax.numpy code
import jax import jaxlib import jax.numpy as jnp print(f"{jax.__version__=}") print(f"{jaxlib.__version__=}") x = jnp.arange(4) y = jnp.arange(20) z =
4
comments
Opened 1 other issue in 1 repository
google/jax
1
open
Answered 8 discussions in 1 repository
google/jax
google/jax
- The reshape in Jax is much slower than Numpy
- The reshape in Jax is much slower than Numpy
- item assignment
- Unstack operation in Jax
- Getting the Hessian vector Product of a Flax NN output
- Problem with Boolean masks indexing by jax jit
- Is there a way to bail out of JIT-ing?
- ConcretizationTypeError when using `and` but no error when using `jnp.logical_and`