* remove the dead code KeyTangentTy
* replace TyRules.make_tangent with TyRules.zero
* removed ad.instantiate_zeros_aval, which was redundant with ad.instantiate_zeros ever since (1) we removed units and (2) we made Zero carry an aval on it
* fix a bug in backward_pass where we instantiated a Zero at the primal type rather than the corresponding tangent type
* fix _f_bwd in test_keyarray_custom_vjp, which had the wrong type (need to return cotangents for all inputs, we were returning a (float_tangent, key_tangent) pair instead of a (float_tangent, (float_tangent, key_tangent)) nested tuple, see #19009 for a check which catches this and hence includes the same test change
We probably also need a TyRules.add for any extended dtypes that can occur as tangent dtypes, but we currently don't have any tests that exercise that (because all extended dtype tangent types are currently float0). I have some follow-up work to add such a case though!
This PR is a follow up to #18881.
The changes were generated by adding
from __future__ import annotations
to the files which did not already have them and running
pyupgrade --py39-plus --keep-percent-format {jax,tests,jaxlib,examples,benchmarks}/**/*.py
Previously, we introduced support for multi-platform lowering, by
adding a new LoweringParameters object that can be used to specify
a cross-lowering platform or even multiple platforms. But we had
kept the ModuleContext.platform in place because some lowering rules
were still referencing it. Now we replace ModuleContext.platform with
ModuleContext.platforms, which removes the redundancy, simplifies
the code, and makes it clearer that the lowering rules should not
simply assume single-platform lowering.
PiperOrigin-RevId: 576575376
Currently JAX callbacks on TPU raise errors when the called function takes empty arguments or returns empty results. It seems that the send_to_host function works
even with empty arrays, but recv_from_host crashes (crash log below).
Here we work around this issue, by ensuring that only the non-empty results of the Python callback are sent to the device computation and the empty results are replaced with empty constants in the device computation.
This is part of the work to replace uses of host_callback with io_callback.
PiperOrigin-RevId: 559061336
This flag was added in https://github.com/google/jax/pull/8678 in December 2021
when we changed the behavior of host_callback to not have special handling for autodiff. Nobody is using that flag now.
This is part of a longer project to replace uses of host_callback with jax.pure_callback and jax.experimental.io_callback.
PiperOrigin-RevId: 557520668
Refactoring only, no changes intended. The goal is to shrink xla.py down to only its HLO-compatibility role, and remove things that aren't related to HLO compatibility.
Remove an unused top_k translation rule as well.
PiperOrigin-RevId: 554946059
Change flags to use the newer definition style where the flag is read via a typed FlagHolder object returned by the DEFINE_... function. The advantage of doing this is that `flag.value` has a type known to the type checker, rather than reading it as an attr out of a gigantic config dictionary.
For jax.config flags, define a typed FlagHolder object that is returned when defining a flag, matching the ABSL API.
Move a number of flags into the file that consumes them. There's no reason we're defining every flag in `config.py`.
This PR does not change the similar "state" objects in `jax.config`. Changing those is for a future PR.
PiperOrigin-RevId: 551604974
Notable changes:
* use PEP 585 type names
* use PEP 604 type union syntax where `from __future__ import annotations` is present.
* use f-strings in more places.
* remove redundant arguments to open().
We have a number of potential use cases where we want different functions that interpret pytrees differently. By allowing multiple pytree registries the same tree node can be registered in registry but not another.
One motivating use case is the new opaque PRNG array type. We want `jit` to treat these objects as if they were pytrees, but we want other transformations to leave them alone or handle them specially.
PiperOrigin-RevId: 549301796
Why? This is generally used for static operations on shapes, but np.prod
has an unfortunate corner-case behavior that np.prod([]) returns a float.
math.prod is available as of Python 3.8, and is a better solution here.
* Move dependencies of sharding_impls into sharding_impls to avoid creating cyclic dependencies.
* Fix a handful of new pytype errors.
PiperOrigin-RevId: 523146076
Outfeed receiver compiles computations (during shutdown), and if the correct options aren't provided, then it may not be able to do things like find ptxas for CUDA builds. Plumb the executable build options through from Python.
PiperOrigin-RevId: 518852909
Limit jax._src.lib to shims around jaxlib and nothing else.
The goal of this change is to avoid a dependency cycle between the rest of jax and jax._src.lib in a Bazel build. This allows the types for jax._src.lib to be inferred by pytype in isolation without referring to the rest of JAX.
PiperOrigin-RevId: 512922397
This CL renames occurrences of "mhlo" in: 1) names, 2) tests, 3) prose in order
to prepare for the upcoming migration.
Unchanged occurrences:
1) Public API that contains "mhlo", e.g. XlaLowering.mhlo and the "mhlo"
argument value in Lowering.as_text and Lowering.compiler_ir.
2) Documentation (changelog, JEPs, IR examples, etc).
3) One rare situation where prose says "StableHLO" and "MHLO" in one sentence,
so both are necessary to disambiguate.
PiperOrigin-RevId: 495771153
