Why it matters
QLoRA theory shapes efficient FT. Understanding shapes design.
Advertisement
The architecture
NF4: normalized float 4 for base.
Double quantization for scales.
Paged optimizers.
LoRA on top in bf16.
Advertisement
How it works end to end
NF4 information-theoretic optimal for normal-distributed weights.
Double quant reduces scale storage.
Paged optim protects OOM.