Add runable script to perform the complete experiment done in the PGDL competition #1
Description
As of 12/28/2024, this repository contains the following:
- A script with functions to compute the margin distribution for a converted model from the PGDL competition
- A script with a class that will base its architecture and weights on a model from the PGDL competition.
- Note that this is not a general tensorflow to pytorch converter , but could be adapted to one pretty easily.
- A script that will, given a path to the code and models from the PGDL competition, convert the models of
task1_v4from Tensorflow to Pytorch.
We need a script to be able use the converted models to replicate the margin distribution and confirm the replications. This will require a new script and an update to the README file.