@delock

deepspeed-finetune guides users through hardware assessment, plan selection, and training launch for large language models using deepspeed optimization, with explicit vram budgeting, activation checkpointing tuning, and multi-gpu support.