laishawadhwa
commented
7 months ago @DavidMertz : While AllenNLP does sound relevant it's a specific stack based on Pytorch. I think adding a deep-learning module on Pytorch (NLP and model training) and then a submodule on AllenNLP would be a good addition. We can shift gears a little, and add information on LLM's and extend the focus on the current state of NLP. We can add hands-on for setting up a custom transformer/GPT to build up on the knowledge shared during the NLP course.
Do we want to include the AllenNLP stuff? Is it too far out of date in general, or is this still a good go-to tool for some NLP work?
I developed it initially for the course series I did with Pearson/O'Reilly Online. In the half dozen or so times I taught the sequence, I went through a 3 part series of beginner/intermediate/advance, each part being 4 hours. I found that I rarely got to any of the AllenNLP stuff, and when I did it was mostly because I skipped something else.
However, when I did that with O'Reilly Online, it was pretty much straight lecture, with frustratingly few student questions, and no time spent on exercises. With Erudio, we definitely want to encourage student interaction and do in-class exercises.