TensorFlow Hub provides large variety of models. This repository helps to go through all TF Lite models in TFHub one-by-one and collect statistics about operations in them. Each model will be kept on the machine only for necessary amount of time to avoid out-of-memory
pip install -r requirements.txt
playwright install chromium
1 step. Scrub all the links to tflite models from TFHub site (may take ~15 minutes)
python collect_links_to_models.py
As a result of the first step you will find a file log.dsv
. It could be read via excel-like pro
2 step.
python download_and_collect.py