Add notes about the new CUDA version restrictions to the changelog an…

…d installation instructions.

CHANGELOG.md

@@ -16,18 +16,24 @@ Remember to align the itemized text with the first line of an item within a list
     * `random.rbg_key(seed)` becomes `random.PRNGKey(seed, impl='rbg')`
     * `random.unsafe_rbg_key(seed)` becomes `random.PRNGKey(seed, impl='unsafe_rbg')`
 * Changes:
-  * {func}`jax.scipy.stats.mode` now returns a 0 count if the mode is taken
-    across a size-0 axis, matching the behavior of `scipy.stats.mode` in SciPy
-    1.11.
+  * CUDA: JAX now verifies that the CUDA libraries it finds are at least as new
+    as the CUDA libraries that JAX was built against. If older libraries are
+    found, JAX raises an exception since that is preferable to mysterious
+    failures and crashes.
   * Removed the "No GPU/TPU" found warning. Instead warn if, on Linux, an
     NVIDIA GPU or a Google TPU are found but not used and `--jax_platforms` was
     not specified.
+  * {func}`jax.scipy.stats.mode` now returns a 0 count if the mode is taken
+    across a size-0 axis, matching the behavior of `scipy.stats.mode` in SciPy
+    1.11.
 
 # jaxlib 0.4.17
 
 * Changes:
   * Python 3.12 wheels were added in this release.
-  * The CUDA 12 wheels now require CUDA 12.2.
+  * The CUDA 12 wheels now require CUDA 12.2 or newer and cuDNN 8.9.4 or newer.
+  * This is likely the last release of jaxlib that supports CUDA 11. We intend
+    to drop CUDA 11 support in the next release.
 
 * Bug fixes:
   * Fixed log spam from ABSL when the JAX CPU backend was initialized.

docs/installation.md

@@ -58,8 +58,9 @@ not being installed alongside `jax`, although `jax` may successfully install
 ### pip installation: GPU (CUDA, installed via pip, easier)
 
 There are two ways to install JAX with NVIDIA GPU support: using CUDA and CUDNN
-installed from pip wheels, and using a self-installed CUDA/CUDNN. We recommend
-installing CUDA and CUDNN using the pip wheels, since it is much easier!
+installed from pip wheels, and using a self-installed CUDA/CUDNN. We 
+strongly recommend installing CUDA and CUDNN using the pip wheels, since it is
+much easier!
 
 JAX supports NVIDIA GPUs that have SM version 5.2 (Maxwell) or newer.
 Note that Kepler-series GPUs are no longer supported by JAX since
@@ -87,6 +88,13 @@ pip install --upgrade "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-re
 pip install --upgrade "jax[cuda11_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
 ```
 
+If JAX detects the wrong version of the CUDA libraries, there are several things
+to check:
+* make sure that `LD_LIBRARY_PATH` is not set, since `LD_LIBRARY_PATH` can
+  override the CUDA libraries.
+* make sure that the CUDA libraries installed are those requested by JAX.
+  Rerunning the installation command above should work.
+
 ### pip installation: GPU (CUDA, installed locally, harder)
 
 If you prefer to use a preinstalled copy of CUDA, you must first
@@ -106,13 +114,13 @@ able to use the
 that NVIDIA provides for this purpose.
 
 JAX currently ships two CUDA wheel variants:
-* CUDA 12.0 and CuDNN 8.9.
+* CUDA 12.2 and CuDNN 8.9.4.
 * CUDA 11.8 and CuDNN 8.6.
 
-You may use a JAX wheel provided the major version of your CUDA and CuDNN
-installation matches, and the minor version is at least as new as the version
-JAX expects. For example, you would be able to use the CUDA 12.0 wheel with
-CUDA 12.1 and CuDNN 8.9.
+You may use a JAX wheel provided the major version of your CUDA
+installation matches, and the minor versions are the same or newer.
+JAX checks the versions of your libraries, and will report an error if they are
+not sufficiently new.
 
 Your CUDA installation must also be new enough to support your GPU. If you have
 an Ada Lovelace (e.g., RTX 4080) or Hopper (e.g., H100) GPU,