• Relevant_Outcome_726@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    10 months ago

    From my experience, Here are some other things related to Lora:
    + FSDP doesn’t work for Lora because FSDP requires all parameters to be trainable or frozen.

    + For Qlora, currently we can only use deepspeed zero2 (deepspeed zero 3 is not supported)