I am trying out the newly added TPU support on Google Colab with the simple cats vs dogs dataset.
After creating a simple CNN, I tried to export the model to TPU. But it failed with error
TypeError: Checkpointable._track_checkpointable() passed type <class 'keras.engine.topology.InputLayer'>, not a Checkpointable.
Here's the code that I wrote on Colab.
model = models.Sequential()
model.add(layers.Conv2D(32, (3,3), activation='relu', input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2,2)))
model.add(layers.Conv2D(64, (3,3), activation='relu'))
model.add(layers.MaxPooling2D((2,2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))
model.summary()
train_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(train_dir, target_size=(150,150), batch_size=20, class_mode='binary')
tpu_model = tf.contrib.tpu.keras_to_tpu_model(model, strategy=tf.contrib.tpu.TPUDistributionStrategy(tf.contrib.cluster_resolver.TPUClusterResolver(tpu="grpc://" + os.environ['COLAB_TPU_ADDR'])))
My guess is I am doing something wrong in train_generator
. But I am not sure what it is. Any help would be highly appreciated.
If you're using or import layers
from Keras
instead of TensorFlow
like this:
from keras import layers,models
from keras.preprocessing.image import ImageDataGenerator
import tensorflow as tf
You will get error like you mention above :
TypeError: Checkpointable._track_checkpointable() passed type <class 'keras.engine.topology.InputLayer'>, not a Checkpointable.
So, you can import layers
directly from TensorFlow
like my code below:
from tensorflow.keras import layers,models
from keras.preprocessing.image import ImageDataGenerator
import tensorflow as tf
or you can see my full code here: https://gist.github.com/ilmimris/8218e397dd35ab693404e95db32dc574
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With