davearcher18 / llama-program

This repository contain the work done while enrolled in the Llama Program at Lambda AI

llama-program

This repository contain the work done while enrolled in the Llama Program at Lambda AI

First project: Producing a Pytorch Lightning callback that plots confusion matrices while training a model.

llama-program's Issues

Investigate the distribution of clout score for different sign up sources.

We have various scouts signing up from different places. We want to know if different sources results in better scouts and you need to provide a report on if it does.

Point of contact: Cobus

Write a generic function to plot every scouts clout score over time

Cobus requires a function to plot the clout score over time. Speak to him about how the data looks like and what exactly he wants.

Point of contact: Cobus

Automate scout morning report.

The script needs to take in the various reporting functions and combine them into one report that gets send out every day.

Point of contact: Cobus

Go through the traffic cropped images dataset and flag images that are incorrect

We are using an object detection model to crop the required vehicles for our model. Go through the newly created cropped dataset and flag images that are not the type of vehicle they are labeled.

Add a notebook for each one of your projects from the first few weeks to the github repo.

Investigate the traffic images and identify images that are labeled incorrectly.

We have received a dataset of roughly 75K images, but after further investigation we found there are various cases where the images are labeled incorrectly.

Please take my subset of data and investigate the images for wrongly labeled data. You do not have to go through everything, but we need a rough idea.

Point of contact: Ruan

Make confusion matrix in callback give more information than simply the amount of numbers. For instance, per row the performance and a constant performance for the diagonal as well.

Add weights and biases logging functionality to all callbacks.

Weights and biases is turning into Lambda-AI primary resource for monitoring models as well as generating reports.

For your current two callbacks, see if you can automatically log the outputs to a weights and biases project page.

Point of contact: Ruan