As we know that the Deep Learning frameworks, consumed in most business operations, are growing rapidly. I have surveyed that Python, Java and C++ are the most widely used languages for Deep Learning. I am exploring TensorFlow (TF 2.x), Apache Spark, Pytorch, MXNet, Sonnet and AWS SageMaker and these Machine Learning frameworks support Deep Learning.
I am trying to find that out of these Deep Learning frameworks, which frameworks support streaming in (Prediction & Training) and Transfer Learning(Incremental Learning)?.
There is one initial Training on Large Dataset, and then multiple frequent trainings on smaller and similar datasets.
Also, can anyone confirm:
1)-If KAFKA is the only option to enable streaming in TF?
and
2)-If Transfer Learning in TF is the only option for retraining the model(training multiple times, using exiting trained model)?