jakevdp
Follow
jakevdp
Follow
Achievements
Achievements
Highlights
- 168 discussions answered
Pinned
1,489 contributions in the last year
Contribution activity
May 2022
Created 33 commits in 1 repository
Created a pull request in google/jax that received 10 comments
DOC: add references for haskell-style signatures
Fixes #8355 See the rendered version e.g. here: https://jax--10700.org.readthedocs.build/en/10700/_autosummary/jax.lax.scan.html
+38
−16
•
10
comments
Opened 32 other pull requests in 1 repository
google/jax
2
open
29
merged
1
closed
- WIP: make lax_numpy_test compatible with strict dtype promotion
- [x64] make jnp.unravel_index safe under strict promotion
- [x64] make jax.random compatible with jax_numpy_dtype_promotion=strict
- [x64] jax.numpy reductions: avoid binary promotion for upcast_bf16
- [x64] handle strict promotion for jnp.var
- [x64] linspace/logspace/geomspace: avoid problematic type promotions
- [x64] dtypes_test: run tests with jax_numpy_type_promotion=strict
- BUG: fix internal _jax_type for weak bfloat16
- jnp.nonzero: require array-like input
- WIP: [x64] set jax_numpy_dtype_promotion='strict' in tests
- [x64] add jax_numpy_dtype_promotion flag
- [x64] use explicit casting rules for promote_dtypes_inexact
- [sparse] test corner cases of sparse.eye
- DOC: fix scipy intersphinx link
- [sparse] add sparse.eye() array creation routine
- DOC: note compilation of funcs in lax control_flow
- [sparse] raise better error for subtraction between sparse and dense
- api_test: fix typo in expected error message
- [sparse] add BCOO.update_layout() method
- Remove unnecessary __future__ imports
- DOC: explicitly enable dollarmath extension
- DeviceArray: Improve support for copy, deepcopy, and pickle
- DOC: update myst-nb to v0.15.0
- DOC: clarify jupytext instructions
- DOC: add FAQ entry on jit-compiling methods
- Some pull requests not shown.
Reviewed 21 pull requests in 1 repository
google/jax
21 pull requests
- fix incorrect f-string format in xmap
- [x64] add jax_numpy_dtype_promotion flag
- ENH _wraps() in jax._src.numpy.util is now returns a generic function
- incorrect link in dpsgd example
- DOC: fix docs generator to group notebook code output
- quick fix for #10750, add checks and todo
- Upgrade legacy Python code
- DeviceArray: Improve support for copy, deepcopy, and pickle
- Include mlir.transforms and mlir.passmanager in jaxlib BUILD.
- replace int with operator.index for lax
- Fix mypy warning
- Avoid generating trivial gathers when reversing array
- Update custom interpreter tutorial
- Small fix in FAQ documentation
- DOC: update myst-nb to v0.15.0
- cosmetic edits in docs
- remove jnp.array case for handling buffers w/ aval=None
- api_util: make shaped_abstractify respect raise_to_shaped
- Add jax.scipy.linalg.funm
- jnp.unravel_index: avoid overflow for large dimension sizes
- [sparse] implement sparse rule for lax.reshape_p
Created an issue in google/jax that received 2 comments
Segfault in jax.scipy.linalg.schur for 16-bit floats
Reported in #10490 >>> import jax.numpy as jnp >>> import jax >>> a = jnp.array([[3,2,1],[4,2,1],[4,4,0]], dtype=jnp.bfloat16) >>> jax.scipy.linalg.s…
2
comments
Opened 1 other issue in 1 repository
google/jax
1
open
Answered 10 discussions in 1 repository
google/jax
google/jax
- Best way to handle variable-length data?
- Cache_Clear()
- How to vectorize functions of a custom pytree?
- How to read "type signatures"?
- sklearn.metrics.pairwise.rbf_kernel is way faster than my JAX implementation
- jax.numpy and implicit conversion?
- Class with jitted function: the following solution is it a good practice?
- Question about memory and computational efficiency for solving large-scale systems of linear equations.
- Using tracers as dictionary keys
- Fixed (Static) Shape and Type for JIT Compilation