3 ms·
Shouldn’t that be “1-trit”?
by transfire 3y ago
Shouldn’t that be “1-trit”?
- QuesnayJr 3y agoThey call it 1.58-bit in the paper. (1.58 is roughly the base 2 logarithm of 3.)
- bmacho 3y agoRead the pdf https://arxiv.org/pdf/2402.17764.pdf https://arxiv.org/pdf/2402.17764.pdf they call it 1-bit everywhere. I don't know why do they do this, 1-bit seems to be a very wrong name for {-1, 0, 1}.
- FrustratedMonky 3y agoYes Technically, but it is catchy for the masses. 1-bit seems to get the idea across, even if not technically describing {-1,0,1}.
- edflsafoiewq 3y agoI think 0 "doesn't count", since you don't have to add or subtract anything for it, just mask it out.
- numpad0 3y agoTernary or three-value logic is a thing in CS[1] 1: https://en.wikipedia.org/wiki/Three-valued_logic https://en.wikipedia.org/wiki/Three-valued_logic
- paipa 3y agoWould be cool to see what happens if you quantize towards zero preferentially. Sparsifying the matrix should improve inference speed directly, right?