- This is an essential python tool to assert the memory requirements of a deep learning model that is implemented utilizing Keras framework and gonna be deployed in TF.
- NN model's size is estimated based on the TF v1.8 specs.
Module requirements:
1. Keras 2.0 or higher
2. Tensorflow 1.8 or higher
(I' ll keep you posted if something change regarding memory TF memory mangement)
How to run
- Just hit: "python model_mem_requirements.py" and you will be prompted!
or
- $ python model_mem_requirements.py -b <batch_size> -p <path/to/model/file> -m <name_of_the_model_definition_in_that_file>
Theory:
total_memory = 4.0*batch_size*(shapes_mem_count + trainable_count + non_trainable_count)
Repo Focus
Similar micro-tools may be provided in the future to constitute a more comprehensive toolkit.