hawkinsp
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 upPopular repositories
-
-
Forked from tensorflow/tensorflow
Computation using data flow graphs for scalable machine learning
C++
-
-
-
1,260 contributions in the last year
Contribution activity
December 1, 2020
hawkinsp has no activity
yet for this period.
November 2020
Created 100 commits in 3 repositories
Created a pull request in google/jax that received 7 comments
Change jaxlib build rules to build a wheel package.
Before this PR, the jaxlib build script (build/build.py) writes its output to build/ subdirectory of the jax source tree. This is a bit ugly since …
+224
−163
•
7
comments
Opened 28 other pull requests in 1 repository
google/jax
1
closed
27
merged
- Remove duplicate code to copy cusolver_kernels in wheel build.
- Update XLA.
- Fix broken CI build due to NumPy 1.18 bug in full_like tests.
- Update XLA.
- Fix integer overflow in np.split on windows.
- Fix lax_reference population_count implementation to avoid overflowin…
- Add np.intc to the set of valid jaxtypes.
- Add -Wno-stringop-truncation to build flags on Linux.
- Fix build.py to work on Linux once again.
- Short-circuit references to jax.core via jax.abstract_arrays.
- Fix mypy error caused by cyclic import dependency.
- Use the jax module to determine the jax source directory base, rather…
- Update bazel version to 3.1.0.
- Relax test tolerance on polymul tests.
- Update XLA.
- Lower complex128 scatter-add to two float64 scatter-adds on GPU.
- Delete jax.nn.functions.
- Add Python 3.9 support to jaxlib build.
- Update XLA.
- Fix manylinux2010 compliance of GPU wheels.
- Fix flake8 warning in jax2tf saved_model_main example.
- Add a deprecation warning to the optix package.
- Fix accidental type promotion in adam and adamax optimizers.
- Add documentation to several functions in jax.lax.linalg.
- Fix max/min confusion in lax.py.
- Some pull requests not shown.
Reviewed 30 pull requests in 2 repositories
google/jax 29 pull requests
- More stable implementation of np.hypot
- readme typo
- jax.numpy: match numpy arguments for reductions
- Build on Windows
- update version & CHANGELOG for jax v0.2.6
- Support shape parameter in jax.numpy.*_like()
- fix mypy error
- Add support for XLA variadic reduce
- Filter out non-integer index_dtypes in argmin/max.
- [jax2tf] Fix mypy error in harnesses.
- Add typing annotations and expose py.typed
- Fix doc for index operations
- [RFC] Support complex dtype in random.normal
- jaxlib: Fix Python 3.9 build and drop CUDA 10.0 support
- Add a FAQ entry on slow JIT compilation
- Remove the private _xla_compilation.
- Fix broken link in readme
- Add symbols for the Python and C++ `DeviceArray` implementation.
- add gpu determinism note
- Cleanup: remove unused utility function
- Fix issue in jax.dtypes._jax_type
- Update weak dtype promotion rules
- Add the non-hashable breaking change to the changelog.
- Change `PyBuffer.shape` to be PyBuffer.xla_shape` in a backward compa…
- Add methods to interact with `DeviceArray` objects.
- Some pull request reviews not shown.
tensorflow/tensorflow 1 pull request
Created an issue in google/jax that received 1 comment
jnp.hypot is numerically unstable
sqrt(x**2 + y**2) is not a good implementation of hypot (e.g., see Wikipedia https://en.wikipedia.org/wiki/Hypot)
1
comment