AttributeError: module 'tensorflow_federated.python.learning' has no attribute 'from_keras_model'

Question

Attempting to run federated learning on tff however, encountering the following:

AttributeError: module 'tensorflow_federated.python.learning' has no attribute 'from_keras_model'

Code: trainer = tff.learning.algorithms.build_weighted_fed_avg( model_fn, client_optimizer_fn=lambda: tf.keras.optimizers.Adam(), server_optimizer_fn=lambda: tf.keras.optimizers.Adam() )

state = trainer.initialize() train_hist = [] for i in range(EPOCHS): state, metrics = trainer.next(state, train_data) train_hist.append(metrics)

print(f"\rRun {i+1}/{EPOCHS}", end="")

Environment: Using Google collab python - 3.10.12 TensorFlow Federated version: 0.61.0

Any help appreciated.

Attempted to downgrade version of TFF

Answer 1

Now "from_keras_model" is moved to "models". So "tff.learning.from_keras_model" is replaced with "tff.learning.models.from_keras_model"

Old Code

  return tff.learning.from_keras_model(
      keras_model,
      input_spec=xxxxx,
      loss=tf.keras.losses.SparseCategoricalCrossentropy(),
      metrics=[tf.keras.metrics.SparseCategoricalAccuracy()])

NEW Code

  return tff.learning.models.from_keras_model(
      keras_model,
      input_spec=xxxxx,
      loss=tf.keras.losses.SparseCategoricalCrossentropy(),
      metrics=[tf.keras.metrics.SparseCategoricalAccuracy()])