Jax 0.4.13 release.

.github/workflows/wheel_win_x64.yml

@@ -1,8 +1,6 @@
 name: Wheel build - Windows CPU x86_64
 on:
   workflow_dispatch: # allows triggering the workflow run manually
-  release:
-    types: [published]
 
 env:
   DISTUTILS_USE_SDK: 1

CHANGELOG.md

@@ -25,7 +25,9 @@ Remember to align the itemized text with the first line of an item within a list
         determine the output shardings.
     * If the mesh context manager is provided, None will imply that the value
       will be replicated on all devices of the mesh.
-    * Executable.cost_analysis() works on Cloud TPU
+  * Executable.cost_analysis() works on Cloud TPU
+  * Added a warning if a non-allowlisted `jaxlib` plugin is in use.
+  * Added `jax.tree_util.tree_leaves_with_path`.
 
 * Bug fixes
   * Fixed incorrect wheel name in CUDA 12 releases (#16362); the correct wheel
@@ -38,9 +40,13 @@ Remember to align the itemized text with the first line of an item within a list
 
 ## jaxlib 0.4.13
 
+* Changes
+  * Added Windows CPU-only wheels to the `jaxlib` Pypi release.
+
 * Bug fixes
   * `__cuda_array_interface__` was broken in previous jaxlib versions and is now
     fixed ({jax-issue}`16440`).
+  * Concurrent CUDA kernel tracing is now enabled by default on NVIDIA GPUs.
 
 ## jax 0.4.12 (June 8, 2023)
 

README.md

@@ -526,10 +526,10 @@ the following in your cloud TPU VM:
 pip install jax[tpu] -f https://storage.googleapis.com/jax-releases/libtpu_releases.html
 ```
 
-For interactive notebook users: Colab TPUs are no longer supported by JAX as of
-version 0.4. However, for an interactive TPU notebook in the cloud, you can use
-[Kaggle TPU notebooks](https://www.kaggle.com/docs/tpu), which are fully
-supported by JAX.
+For interactive notebook users: Colab TPUs no longer support JAX as of
+JAX version 0.4. However, for an interactive TPU notebook in the cloud, you can
+use [Kaggle TPU notebooks](https://www.kaggle.com/docs/tpu), which fully
+support JAX.
 
 ### pip installation: Apple GPUs
 

WORKSPACE

@@ -7,10 +7,10 @@ load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
 #    and update the sha256 with the result.
 http_archive(
     name = "xla",
-    sha256 = "f8f6efd93237c94ee3c8be65dd0702a065c3e3e7b0f732181e62d33813f1bd35",
-    strip_prefix = "xla-34521d16cd91423a041be86bcb35319dfee3bedb",
+    sha256 = "4ec16aff3862c5a243db956ce558d7a62eb79f5e20747b0e80802a3b0d12e419",
+    strip_prefix = "xla-12de6ec958419b57be248d0acd2d9f757e71748c",
     urls = [
-        "https://github.com/openxla/xla/archive/34521d16cd91423a041be86bcb35319dfee3bedb.tar.gz",
+        "https://github.com/openxla/xla/archive/12de6ec958419b57be248d0acd2d9f757e71748c.tar.gz",
     ],
 )
 

setup.py

@@ -19,13 +19,13 @@
 
 from setuptools import setup, find_packages
 
-_current_jaxlib_version = '0.4.12'
+_current_jaxlib_version = '0.4.13'
 # The following should be updated with each new jaxlib release.
 _latest_jaxlib_version_on_pypi = '0.4.12'
 _available_cuda11_cudnn_versions = ['86']
 _default_cuda11_cudnn_version = '86'
 _default_cuda12_cudnn_version = '89'
-_libtpu_version = '0.1.dev20230608'
+_libtpu_version = '0.1.dev20230622'
 
 _dct = {}
 with open('jax/version.py', encoding='utf-8') as f: