Meeting 5

Deep Learning Community of Practice

beginner
code
Deep Learning
Author

Daniel Kick

Published

August 6, 2024

Recap:

This supplementary presentation from this week can be found here.

  • Logistics
    • We’re running a new scheduling poll for the fall since we have members whose schedule may change with the new semester.
    • Because of this, our next meeting will be on the 20th and the one after that will be TBD.
  • MLP in PyTorch
    • We discussed some of the logistics around building models in PyTorch including
    • Dataloaders
    • Moving tensors to the GPU
    • nn.Module classes

Preparation for Next Session

The content for the next two sessions will dovetail:

We’re going to look at transformers next so

  • For next session:

    • Watch Grant Sanderson’s Attention in transformers (26m)

    • Begin Let’s build GPT (1h56)

    • We’re planning to review Santurkar, et al. 2018 (“How Does Batch Normalization Help Optimization?”) for the folks that couldn’t be at Meeting 4. If you haven’t read it give it a read!

    • Decide if you’re game to present a project or paper “show and tell” and what project you’re planning or working on that you.

  • For the session after that:

Cheers,

Daniel