2 ms·
Nice! Would be nice to get various 16-bit float formats as well. They're getting more and more common in graphics and NN domains. Or perhaps generic n-width e
by vardump 7y ago
Nice!
Would be nice to get various 16-bit float formats as well. They're getting more and more common in graphics and NN domains.
Or perhaps generic n-width exponent and m-width mantissa.
- thomasballinger 7y agorreusser made a more generic version inspired by Float Toy: https://observablehq.com/@rreusser/binary-input https://observablehq.com/@rreusser/binary-input
- dunham 7y agoThere is also an 8-bit float used in Lucene. I was surprised that someone was using an 8-bit float format when I first came across that.