Implement overfit test for models

[LukeWood]

Something like:

from keras_cv.models import MLPMixerB16
from tensorflow import keras
import tensorflow as tf 

# constants
batch_size = 10
height = width = 224
num_channels = 3

# model
mixer = MLPMixerB16(include_rescaling=True, include_top=True, classes=10)
mixer.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"])

# data
random_images = tf.random.normal((batch_size, height, width, num_channels))
random_labels = tf.random.uniform((batch_size, ), 0, 5, dtype="int32")

# train
mixer.fit(
   random_images, random_labels, 
   validation_data=(random_images, random_labels), 
   epochs=3
)

Originally posted by @sayakpaul in https://github.com/keras-team/keras-cv/issues/529#issuecomment-1179463713

To make sure we can overfit a randomly generated dataset

[sayakpaul]

@LukeWood assign me this issue.

[LukeWood]

Assigned! Thank you Sayak!

[qlzh727]

Since this involves actual training, which might be too heavy for a unit test. Should we reconsider whether this is needed?

[sachinprasadhs]

@sayakpaul , Considering the above comment, please clarify if we are still going ahead with this implementation.

[github-actions[bot]]

This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you.

[github-actions[bot]]

This issue was closed because it has been inactive for 28 days. Please reopen if you'd like to work on this further.