Notebook 06: TensorFlow API Updates

[mrdbourke]

Work-in-progress to show some of the newer TensorFlow API namespaces.

Functionality remains mostly the same across versions, however, sometimes the namespace of where things are stored gets changed.

Notebook 06 - Transfer Learning with TensorFlow Part 3: Scaling-up

Link: https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/06_transfer_learning_in_tensorflow_part_3_scaling_up.ipynb

EfficientNet Model

Functionality remains the same, however, the import name of EfficientNets in TensorFlow has changed slightly.

import tensorflow as tf

# New (notice the new .efficientnet), see: https://www.tensorflow.org/api_docs/python/tf/keras/applications/efficientnet/EfficientNetB0
model = tf.keras.applications.efficientnet.EfficientNetB0()

# Old
model = tf.keras.applications.EfficientNetB0()

Data augmentation

• from tensorflow.keras.layers.experimental import preprocessing -> from tensorflower.keras import layers

New

from tensorflow.keras import layers
from tensorflow.keras.models import Sequential

# NEW: Newer versions of TensorFlow (2.10+) can use the tensorflow.keras.layers API directly for data augmentation
data_augmentation = Sequential([
  layers.RandomFlip("horizontal"),
  layers.RandomRotation(0.2),
  layers.RandomZoom(0.2),
  layers.RandomHeight(0.2),
  layers.RandomWidth(0.2),
  # preprocessing.Rescaling(1./255) # keep for ResNet50V2, remove for EfficientNetB0
], name ="data_augmentation")

Old

from tensorflow.keras.layers.experimental import preprocessing

# Build data augmentation layer
data_augmentation = Sequential([
  preprocessing.RandomFlip('horizontal'),
  preprocessing.RandomHeight(0.2),
  preprocessing.RandomWidth(0.2),
  preprocessing.RandomZoom(0.2),
  preprocessing.RandomRotation(0.2),
  # preprocessing.Rescaling(1./255) # keep for ResNet50V2, remove for EfficientNet                 
], name="data_augmentation")