search

DavidUdell / sparse_circuit_discovery

Circuit discovery in GPT-2 small, using sparse autoencoding
MIT License
7 stars 1 forks source link

Programmatic hyperlinks to neuronpedia in causal graph HTML for dim names #113

Closed DavidUdell closed 3 weeks ago

DavidUdell commented 1 month ago

There's more broadly a fork in the road as far as frontend design philosophies goes here. One path is printing everything out on physical paper at a scale appropriate to mark-up and annotate that paper in ink. Another path is to write fancy HTML that will be preserved in a .svg file, to be examined entirely in a browser, with any annotations baked in with iframes (h/t Hardik) or hyperlinks.

DavidUdell commented 3 weeks ago

What I'm actually doing is talking to the Neuronpedia api and loading the annotation data into each graph node myself, with my own HTML.