
How TatuEngine stopped blowing its memory budget with ternary weights
A technical post-mortem of how we replaced TatuEngine's weight format with ternary blocks and got the model back inside consumer GPU memory, without losing inference coherence.
3 posts

A technical post-mortem of how we replaced TatuEngine's weight format with ternary blocks and got the model back inside consumer GPU memory, without losing inference coherence.

I actually read the BitNet b1.58 paper and got stuck on a math problem that refused to add up: how do three values (-1, 0, +1) fit in 1.58 bits? A study week that became a didactic clone of ternary quantization — and wrecked my mental model of 'precision equals quality'.

BitMamba-2 1B's in-engine inference representation was unworkable in memory. A compact ternary block codec delivered orders of magnitude of savings, with the lm_head diverging mid-validation.