W2/6 D2/5

Created 2024-04-02 / Edited 2024-04-02

Applied LLM Demos; basic token output visualization with graphviz

Basic token tree

  • General Status:
    • At the hub today!
    • Delighted to keep meeting and chatting with people, went to lunch and all that
    • I organized an Applied LLM meetup -- there were demos that were DELIGHTFUL! A lovely enthusiastic group :)
  • Nerd Snipe:
    • Flipped my bluetooth mouse (which I hadn't used since switching to Sway) to left-handed
    • My conference room presenting connection on this laptop has thrown me; Sway/Wayland won't let you have merged screens and the jabber speaker/mic doesn't output any sound for me! I need to do a more controlled practice run, the one I did wasn't good enough
  • Alloy+LLM+Rails Project:
    • I got initial graphviz output!
    • I then decided that it was too slow to generate, so started to get my intel GPU to work
    • I ran into this exact same issue and gave up
    • Then I started figuring out how to spend money renting someone else's computer
    • ... now I have to be careful. I spent $4.50 pretty quickly. I have a budget for this, but I want to spend that on actual computation instead of forgetting to turn something off and letting it run all night or all weekend for now gain
    • Unfortunately the output didn't work for my first larger run; the nodes in graphviz didn't have the full path and overwrote each other
    • Also I think that the local one can go a lot faster if I don't fully reset it every time; there are savepoints and some other things that might work
    • Still -- it's close to being GUI-able!