jakevdp
Follow
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
Pinned
1,590 contributions in the last year
Contribution activity
July 2020
Created a pull request in google/jax that received 8 comments
Properly set X64 flag in github actions
Trying to confirm my suspicions that our github CI is not actually running any x64 tests... Edit: turns out it isn't (see discussion below). The fi…
+4
−4
•
8
comments
- WIP: exploring weakly-dtyped device arrays
- Improve dtype & test coverage for jnp.fmod
- Cleanup: pass function name rather than function object
- Add jnp.modf(); improve dtype & test coverage for related functions
- Cleanup: canonicalize several dtypes to prevent noisy warnings
- Improve searchsorted implementation
- jnp.linspace & friends: more carefully handle dtypes
- Cleanup test names in random_test.py
- jax.random: use correct x32/x64 default dtypes.
- Tweak Dockerfile to prevent build failure and add TODO
- Improve test coverage for jax.numpy sorting algorithms
- update README for new jaxlib version
- update jaxlib version and changelog for pypi
- update version and changelog for pypi
- Cleanup: fix type issues in lax_numpy.py
- implement jax.numpy.lexsort
- Fix type mismatch in jet rule for abs
- [x64 deprecation] Create _np_array utility routine
- WIP: x64: make Jax default dtypes 32-bit
- Change onp/np to np/jnp in docs & notebooks
- Cleanup: convert uses of 'import numpy as onp' in tests
- Cleanup: convert uses of `import numpy as onp` in library code
- lax_numpy: rename arguments to match numpy
- Fix compilation bug in histogram_bin_edges
- Remove unused private helper function
- Some pull requests not shown.
- Add jnp.modf(); improve dtype & test coverage for related functions
- refine population_count type check
- Replicating sort_complex functionality from np.sort_complex to jax.numpy
- Add erfcx
- jax.random: use correct x32/x64 default dtypes.
- Enable int{8,16} and uint{8,16} tests in lax_test and lax_numpy_test.
- DeviceArray.__iter__ returns DeviceArrays, without host sync
- Address nan issue in zeta
- tweak jnp.repeat not to use jnp on shapes
- Fix `jax.image._resize` function
- add clarification about jit inside indexing error message
- Error message and docstring updates RE: dynamic_slice
- Note in pmap docs that pmap compiles like jit.
- Don't move batch dimensions to start in jnp.einsum.
- Relax test tolerance for conv_general_dilated gradients.
- Use keyword arguments in einsum.
- Update scipy.ndimage.map_coordinates docstring
- Fix bug in jnp.repeat where for array repeats with zero entries (e.g. [1,0,3]) and axis=0 the zero entry was not skipped).
- Fix low probability (4/1000) flake of batching_test on GPU.
- A compilable version of np.repeat.
- Make jnp.take work for empty slices of empty arrays.
- Adding np.sort_complex within lax_numpy.py
- Add pshuffle to docs
- Improve np.intersect1d
- Implement complex convolutions on CPU and GPU.
- Some pull request reviews not shown.
Created an issue in google/jax that received 2 comments
lax silently zeros large integers outside x64 mode
Repro, with jax_enable_x64=False:
>>> from jax import lax
>>> lax.neg(1 << 32)
DeviceArray(0, dtype=int32)
If the input is too large for int64, it …
2
comments