Given a Keras model and the shape of the input data, limit_finder.py
will calculate a lower limit for the memory usage of the model.
python3 limit_finder.py <keras_model> [--n_inputs N_INPUTS] [--input_size INPUT_SIZE] [--input_dtype INPUT_DTYPE] [--verbose]
Where:
keras_model
is the path to a Keras modelN_INPUTS
is an integer indicating the number of images to be processed 'at the same time'INPUT_SIZE
is a string matching<height>x<width>x3
if RGB,<height>x<width>
if grayscale. It's actually a regular expression matching^[0-9]+x[0-9]+x3{0,1}$
INPUT_DTYPE
is a string among {float32, float16, float64}
Note: the calculation of the memory used by the net's input works only in case the input is made of images.