Tag: 4-bit quantisation
-
QLoRA Explained: A 4-Bit Base, NF4, and What It Really Costs
QLoRA puts a 7B to 30B fine-tune on one card. NF4, double quantisation, paged optimizers, the throughput tax nobody mentions, and the…
QLoRA puts a 7B to 30B fine-tune on one card. NF4, double quantisation, paged optimizers, the throughput tax nobody mentions, and the…