• ryedaft
    link
    fedilink
    English
    arrow-up
    1
    ·
    7 days ago

    How on earth would you distribute the model for inference without the weights? The gradients are obviously gone so you can’t continue training on the model. Maybe you can still do some kind of LORA?