LoRA-TMLR-2024/openwebmath-lora-rank-64-20B-tokens
Updated
Continued Pretraining - Math (OpenWebMath)
Model and LoRA adapter checkpoints for Llama-2-7B trained on OpenWebMath for up to 20 billion tokens
Model and LoRA adapter checkpoints for Llama-2-7B trained on OpenWebMath for up to 20 billion tokens