Move Defaults to to end of arg docstring and standardise values

[SamuelMarks]

I've written a Python compiler that can do a bunch of thing. Relevant to Keras I am turning TensorFlow into SQL tables by way of SQLalchemy:

# TensorFlow 2.12.0
$ python3 -m cdd exmod --module 'tensorflow.keras.optimizers' \
                       --emit 'sqlalchemy_table' \
                       -o '/tmp/tf_optimizers'

After deduplicating and running black here are the contents of sgd.py:

"""
Generated from keras.optimizers.sgd.
"""

SGD = Table(
    "SGD",
    metadata,
    Column(
        "learning_rate",
        Float,
        comment="A `Tensor`, floating point value, or a schedule that is a `tf.keras.optimizers.schedules.LearningRateSchedule`, or a callable that takes no arguments and returns the actual value to use. The learning rate",
        default=0.001,
        nullable=False,
    ),
    Column(
        "momentum",
        String,
        comment="float hyperparameter >= 0 that accelerates gradient descent in the relevant direction and dampens oscillations.e., vanilla gradient descent",
        default="0, i",
        nullable=False,
    ),
    Column(
        "nesterov",
        Boolean,
        comment="boolean. Whether to apply Nesterov momentum",
        default=False,
        nullable=False,
    ),
    Column("clipnorm", default=None, nullable=True),
    Column("clipvalue", default=None, nullable=True),
    Column("use_ema", Boolean, default=False, nullable=False),
    Column("ema_overwrite_frequency", default=None, nullable=True),
    Column("weight_decay", default=None, nullable=True),
    Column("global_clipnorm", default=None, nullable=True),
    Column("ema_momentum", Float, default=0.99, nullable=False),
    Column("jit_compile", Boolean, default=True, nullable=False),
    Column("name", String, default="SGD", nullable=False),
    Column("id", Integer, primary_key=True, server_default=Identity()),
)

__all__ = ["SGD"]

As you can see, momentum has an incorrect value because of how you've written your docstrings:
https://github.com/keras-team/keras/blob/0f8e81f/keras/optimizers/sgd.py#L62

Some commentary on the Defaults to syntax as understood by Sphinx:

• https://sphinx-rtd-tutorial.readthedocs.io/en/latest/docstrings.html#the-sphinx-docstring-format
• https://sphinxcontrib-napoleon.readthedocs.io/en/latest/example_google.html

---

So this PR rewrite your docstrings to put the Defaults to at the end and limit it to one value or a boolean condition.

PS: I also noticed inconsistent use of backticks. Did you want them everywhere? - This would disambiguate floats from integers I guess

[qlzh727]

Thanks for the PR. Since this is quite large at the moment, is it possible to split this into smaller ones (one PR per package), so that we can speed up the review (sending to multiple team member.)?

[SamuelMarks]

@qlzh727 I can, in the past I had a script to turn one PR into one per file and that got through all the hoops until they reached TensorFlow master.

So you want me to do that here? - Happy to. - Maybe I'll link back here in case others want to chime in about repository-wide standards? - Also if any dis/agreements of this approach were to be made, maybe make them on this issue #17748

EDIT: Here's my script, was hitting GraphQL rate limits so added a sleep:

# Setup branch, commit, push up branch
while read -r name; do (
    branch="${name//\//.}"; branch="${branch::-3}-defaults-to";
    git checkout master && \
    git checkout -b "${branch}" && \
    git checkout defaults-to "$name" && \
    git commit -S -m '['"${name}"'] Standardise docstring usage of "Defaults to"' && \
    git push --set-upstream mine "$branch"; 
) done< <(git diff --name-only)

# Send PRs after `git diff --name-only > /tmp/names` on soft-reset `defaults-to` branch
while read -r name; do
    branch="${name//\//.}"; branch="${branch::-3}-defaults-to";
    git checkout "$branch" && \
    gh pr create \
        --title '['"${name}"'] Standardise docstring usage of "Default to"' \
        --body 'This is one of many PRs. Discussion + request to split into multiple PRs @ #17748' ; 
    sleep 5m ; # was doing `sleep 1h` before
done< <(cat /tmp/names)