llama-cookbook

feat: Add activation checkpointing support for single GPU training

#971

Open