Tagged: bitsandbytes

5 articles

Latest

LLM.int8() Cuts BLOOM-176B Memory in Half—No Performance Loss

QLoRA squeezes 65B-parameter finetuning onto a single 48GB GPU

Hugging Face: bitsandbytes wins on ease, GPTQ wins on speed

New Diffusers quantization slashes Flux GPU memory from 31GB to under 18GB

GaLore Cuts Optimizer Memory 82.5%, Fits 7B Models on an RTX 4090