TX · Utils

Mist Box LLM Training Reference

Open language models sized for my Mist box — RTX 3070 (~8 GB VRAM), 94 GB RAM, 16 cores. All estimates assume BF16 and include training memory overhead (not weights alone).

mist.train.fit()

Training fit on 8 GB

Full = AdamW + BF16 mixed + grad checkpoint (~16 B/param). LoRA = BF16 base (~2 B/param) + ~1.8 GB adapter overhead. QLoRA = NF4 base (~0.5 B/param) + same overhead. Usable budget: 7.5 GB.

Model Params Full LoRA QLoRA Notes
Loading…