Training Startup
Call tf.estimator.train_and_evaluate to train a model.
For details, see "Training" in Porting with Estimator.
When train_and_evaluate is executed in the Estimator scenario, if dynamic capacity expansion of the on-chip memory is not enabled, a table is created twice. For large tables, this may lead to insufficient device memory. To avoid this issue, enable dynamic capacity expansion of the on-chip memory, which ensures that the table is created only once.
Parent topic: Estimator Training