keras image classification model using AL

[noy2121]

Hi, I have a few questions:

1. When performing initial training via ActiveLearner(estimator=keras_classifier, X_training=X_train, y_training=y_train)
   does X_training, y_training support generators?

2. Working with HD images, I need to save huge amounts of data in X_pool which kills the process. Is there a method that takes a subset of X_pool for each query iteration? maybe learner.teach(X_pool=X_pool) also supports generators?

3. The learner only gets NumPy array object or does it supports tf.tensors?

thank you!