[AIR] Allow user to pass model to TensorflowCheckpoint.get_model

[bveeramani]

Signed-off-by: Balaji Veeramani balaji@anyscale.com

Why are these changes needed?

When you resume training a TensorFlow model, you may need to create an unnecessary lambda to load model weights. This is because TensorflowCheckplint.get_model expects a model definition, but you may have already constructed your model. This PR improves the UX by letting users directly pass a model to get_model.

Before

def train_loop_per_worker():
    ...

    model = tf.keras.applications.MobileNetV3Small(input_shape=(3, 260, 260))
    if checkpoint is not None:
        model = checkpoint.get_model(lambda: model)

After

def train_loop_per_worker():
    ...

    model = tf.keras.applications.MobileNetV3Small(input_shape=(3, 260, 260))
    if checkpoint is not None:
        model = checkpoint.get_model(model)

Related issue number

Closes #31290

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[amogkam]

Nice! left some small comments.