*
   * @param sets the sets to choose elements from, in the order that the elements chosen from those
   *     sets should appear in the resulting lists
   * @param <B> any common base class shared by all axes (often just {@link Object})
   * @return the Cartesian product, as an immutable set containing immutable lists
   * @throws NullPointerException if {@code sets}, any one of the {@code sets}, or any element of a

  `Oh, as to the whiting,' said the Mock Turtle, `they--you've
seen them, of course?'

  `Yes,' said Alice, `I've often seen them at dinn--' she
checked herself hastily.

  `I don't know where Dinn may be,' said the Mock Turtle, `but
if you've seen them so often, of course you know what they're
like.'

  `I believe so,' Alice replied thoughtfully.  `They have their

// "/job:localhost/replica:0/task:0/device:CUSTOM:0".
//
// The custom device defines copy operations for moving TensorHandles on and
// off, and an execution operation for named operations. Often execution will
// simply wrap op execution on one or more physical devices.
//
// device_info is an opaque caller-defined type stored with the custom device

3. Copy the file with the requirements to the `/code` directory.

    Copy **only** the file with the requirements first, not the rest of the code.

    As this file **doesn't change often**, Docker will detect it and use the **cache** for this step, enabling the cache for the next step too.

4. Install the package dependencies in the requirements file.

   *
   * @param sets the sets to choose elements from, in the order that the elements chosen from those
   *     sets should appear in the resulting lists
   * @param <B> any common base class shared by all axes (often just {@link Object})
   * @return the Cartesian product, as an immutable set containing immutable lists
   * @throws NullPointerException if {@code sets}, any one of the {@code sets}, or any element of a

index with a random number of shards: there should be one test for the 1-shard
case, and another test for the 2+ shards case.

==== Abuse randomization in multi-threaded tests

Multi-threaded tests are often not reproducible due to the fact that there is
no guarantee on the order in which operations occur across threads. Adding
randomization to the mix usually makes things worse and should be done with
care.

  `Oh, as to the whiting,' said the Mock Turtle, `they--you've
seen them, of course?'

  `Yes,' said Alice, `I've often seen them at dinn--' she
checked herself hastily.

  `I don't know where Dinn may be,' said the Mock Turtle, `but
if you've seen them so often, of course you know what they're
like.'

  `I believe so,' Alice replied thoughtfully.  `They have their

   * name within that type. The constant name should be a sensible identifier that is closest to the
   * "common name" of the media. This is often, but not necessarily the same as the subtype.
   *
   * Be sure to declare all constants with the type and subtype in all lowercase. For types that

common:rbe_cross_compile_macos_x86 --bes_upload_mode=nowait_for_upload_complete
test:rbe_cross_compile_macos_x86 --config=rbe_cross_compile_base
# Increase the test timeout as tests often take longer on mac.
test:rbe_cross_compile_macos_x86 --test_timeout=300,450,1200,3600
# Limit jobs to 100 to avoid running into "out of memory" issues (b/316266643)
common:rbe_cross_compile_macos_x86 --jobs=100

   * query than the original; your mileage may vary. Precomputation takes time and requires more
   * memory, so it is only likely to be worthwhile if the precomputed matcher is queried very often.
   *
   * <p>This method has no effect (returns {@code this}) when called in GWT: it's unclear whether a
   * precomputed matcher is faster, but it certainly would consume more memory (which doesn't seem