TL;DR

I downloaded some datasets and got Facebook Meta AI's MDETR model running.

Today's Research Goals

• Actually start taining the MDETR model.

Completed Tasks

• I was able to run a script to start the MDETR model training.
• I extracted noun phrases from the captions of the Flickr30k dataset.

Here's an image of some noun phrases:

![A mapping of image names to noun phrases in their captions.]("./assets/Noun Phrases!.png")

Unresolved Things

[None]

Questions From Today

• What do these benchmarks measure, exactly?
• Why is there no standard for code for ML models?

Other Notes

I am very much digging this mechanical keyboard provided by my lab. Before joining this lab, I could not see what the fuss was about. Now, I think I'll purchase a mechanical keyboard of my own once I have the funds (and a dedicated desktop of my own).

I have also forgotten just how powerful it is to have a machine that accepts inputs faster than you can type and that does not crash every few hours. I want a new laptop.

Also, remind me never to copy-paste a configuration file from the web into Vim again.

![It's been going for about an hour.]("./assets/And They Don't Stop Coming.png")