Closed czq693497091 closed 1 month ago
@czq693497091 Thanks for your interest! The main branch for FlexLLM and peft-related functionalities is peft
. The other branches will be merged into the peft
branch in the next few weeks. An old version of token-level fine-tuning is available in this branch; we will upstream the newest version into the peft
branch at a later time (likely a month or two).
Hi, I am really interested in the work of FlexLLM and I guess the artifact evaluation is realeted to the peft_* branch. I wonder the difference between the peft, peft-qkv and peft_optimizer. Especially I am impressed by the design of token-level finetune. Which branch is recommended to study the FlexLLM? Thanks!