variint.bsky.social
Mechanistic interpretability of modular convnets applied to 👁️ at U Miami and 🛰️🐝 at Boku Vienna | she/her 🦒💕
variint.github.io
105 posts
1,901 followers
758 following
Regular Contributor
Active Commenter
comment in response to
post
I think we are a bit stuck there, attention gates can't be the last invention in the field. I feel like everyone is so focused on it that people forget to invent new things (?) 👀 I guess it's a matter of time!
comment in response to
post
Not sure, but no 👀😂
comment in response to
post
My goal was actually to create the quiz for studying (similar to Duolingo) - after the 3rd time you should be able to answer all the questions 😀 Thank you, but I stole it - I had the link below my other post 😇 - (codepen.io/Pedro-Ondivi...)
comment in response to
post
Inspired by: codepen.io/Pedro-Ondivi...
comment in response to
post
currently Python (with PyTorch) ☺️
comment in response to
post
@europlanetmedia.bsky.social 🛰️
comment in response to
post
Seems like you were more productive 🥲
comment in response to
post
My prof has a github repo, and I am trying to extend it a little😊 (github.com/harbourlab/S...)
comment in response to
post
I think I know why now.
comment in response to
post
Welcome!! 🦋
comment in response to
post
I love that challenge!!
comment in response to
post
This is such a cool idea, I love it!! Also, how many hours does this take? Seems like a lot 😂
comment in response to
post
I want to work on vision models that can run on my laptop with small datasets. I have two applications, one is retinal disease understanding and one focuses on remote sensing imaging. The goal is to align concepts (e.g. "lesion" in medical image data) to a part of the model 😊
comment in response to
post
Ahh, yes, I guess all the context is on that other platform we all recently left 😅 I am working with convolutional neural networks - trying to cut some connections (sparsifying/pruning) to make the model more explainable.
comment in response to
post
Thanks 😊
comment in response to
post
I think that's the page most people use to find them 😊 blueskydirectory.com/starter-pack...
comment in response to
post
There are tons of starter packs around, you can follow them/individual people and ask the creators to be part of it. Also, there are block lists - you can block whole lists of people, many focus on hate speech.
comment in response to
post
Love it!!
comment in response to
post
Wohoo, progress!! And there is always too much coffee 😂
comment in response to
post
so beautifuuul, I love it
comment in response to
post
I am trying to create a node-distance loss function for a deep neural network 😃