Implement Machine Learning Experiments - Pytorch -- 2

Hi! I'm a machine learning researcher looking to hire someone to help me implement some research experiments. The end goal is to publish the results of these experiments in an academic conference or journal. I am capable of implementing these things myself. However, I'm interested in finding out how much this work would cost and how much time it will save me. This project will be very short, and is just a test run to see if I would be interested in using a service like this again. The experiment involves training a model, pruning it using a provided library, and then continue training the model. Specifically, you will implement several un-implemented python functions which: - Train VGG-16 on CIFAR-10 (with provided hyper-parameter settings) and test the development accuracy - Prune a VGG-16 model using a provided library - Evaluate the development accuracy of a pruned model - Evaluate the distance between the hidden activations of a pruned and an original model - Continue training a pruned model You will not actually need to run any of these scripts (so you do not need GPU access). You will just implement the interface, and I will take care of training the models and combining the scripts using bash. These tasks are already implemented for learning MNIST using a simple feed-forward neural network, so I expect this will be fairly simple to do if you use the existing functions as a reference. Attached is an example function which trains a 2-layer feed-forward neural network on MNIST, and another example which prunes that network using a provided library.