Ultra-low VRAM matrix-factorized PyTorch optimizer. Slashes AdamW memory by 99.99% down to a constant ~MB footprint across any model scale.
training optimize optimizer pytorch optimized artificial-intelligence artificial-neural-networks artificial-intelligence-algorithms pytorch-implmention factorized pytorch-implementation optimizers factorized-networks factorized-models optimizer-tool
-
Updated
Oct 31, 2026 - Python