SOURCE RECORD · github
Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.
Open canonical source ↗
Research paper status → · Book status →