Your training script must be configured to write TensorBoard logs. For existing TensorBoard users, this requires no change to your model training code.
To configure your training script in TensorFlow 2.x, create a
TensorBoard callback and set the log_dir
variable to any location
which can connect to Google Cloud.
The TensorBoard callback is then included in the TensorFlow model.fit
callbacks list.
import tensorflow as tf
def train_tensorflow_model_with_tensorboard(log_dir):
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
def create_model():
return tf.keras.models.Sequential(
[
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(512, activation="relu"),
]
)
model = create_model()
model.compile(
optimizer="adam",
loss="sparse_categorical_crossentropy",
metrics=["accuracy"]
)
tensorboard_callback = tf.keras.callbacks.TensorBoard(
log_dir=log_dir,
histogram_freq=1
)
model.fit(
x=x_train,
y=y_train,
epochs=5,
validation_data=(x_test, y_test),
callbacks=[tensorboard_callback],
)
The TensorBoard logs are created in the specified directory and can be uploaded to a Vertex AI TensorBoard experiment by following the Upload TensorBoard Logs instructions for uploading.
For more examples, see the TensorBoard open source docs
What's next
- Check out automatic log streaming