正在加载视频...

视频加载失败

what if visualizing your data was actually a video game? flying through 30 million jina-v5-nano embeddings of 12 billion tokens of multilingual FineWeb + starcoder + pile + RedPajama looking at your data is supposed to be like eating your veggies, what if we made tasty veggies?

23,261 次观看 • 23 天前 •via X (Twitter)

14 条评论

Ian Johnson 🔬🤖 的头像
Ian Johnson 🔬🤖23 天前

each voxel is a bin that holds umap data points, so we can preview whats in it with a heatmap (to show density), summary statistics (data set composition) or experimenting with an identicon (icon generated from content hash). i have more ideas but the weekend is over

Smoke-away 的头像
Smoke-away23 天前

Heatmap view is awesome.

Daniel van Strien 的头像
Daniel van Strien22 天前

You should make this into a @huggingface Space (even cooler if you can use Jobs so someone can set this up for any hub dataset!)

Vorname MitD 的头像
Vorname MitD22 天前

Veggies attract fruit flies - I see a pattern here :)

William Wolf 的头像
William Wolf23 天前

so cool

lakshya 的头像
lakshya23 天前

so cool

Kushal 的头像
Kushal22 天前

Severance would like a call

Caleb Palmquist 的头像
Caleb Palmquist22 天前

This is very cool, I’ve thought of doing something like this before. Are there some practical applications you are using this with?

Benny Schudel 的头像
Benny Schudel22 天前

This is amazing ❤️

Dr. Xi Zeng 的头像
Dr. Xi Zeng22 天前

I'd go looking for the single-point voxels you filtered out. Those samples might be more fun to explore than the dense neighborhoods.

Matt Caldwell 的头像
Matt Caldwell22 天前

mindcraft

Alex 的头像
Alex23 天前

30 million embeddings is the scale where 'explore vs. report' stops being a UX question and starts being an epistemology one

Thomas Czerniawski 的头像
Thomas Czerniawski23 天前

How many voxels?

Ian Johnson 🔬🤖 的头像
Ian Johnson 🔬🤖23 天前

3 million. but i filtered the view to only voxels with >5 points and its 500k (1.5m have just 1 point in them)

相关视频