cuda

3 posts

Aug 12, 2026

TatuEngine: the codec that shrank a 27GB model into 245MB — and the lm_head that wouldn't cooperate

The BitMamba-2 1B's 1:1 BVH took 27.26GB — unworkable. A compact ternary block codec delivered 114× compression, with the lm_head diverging mid-validation.

Continue reading →

tatuengine compression bvh

TatuEngine

Aug 03, 2026

The story of TatuEngine — from dream RT Cores to an autopoietic agent

Born from the absurd question 'what if we used RT Cores to process neural networks?', killed the dream on WSL2, and became a hybrid CPU/GPU engine with BitMamba-2 1B, radical compression, an agent with 8 subsystems and a security sandbox — 302 commits later.

BitMamba 1B — training the first SSM model

The journey (and the craters) of training a Mamba-2 with 1.4B parameters from scratch on an RTX 3060 — warmup, SFT, catastrophic eval, and Phase 4 of the resurrection.

Continue reading →

tatuengine bitmamba training

Browse by tags

TatuEngine: the codec that shrank a 27GB model into 245MB — and the lm_head that wouldn't cooperate

The story of TatuEngine — from dream RT Cores to an autopoietic agent

BitMamba 1B — training the first SSM model