• keepthepace@slrpnk.net
    link
    fedilink
    English
    arrow-up
    3
    ·
    11 months ago

    Nice! It feels like a direct answer to Karpathy comment on Mistral, where he said it is nice to call it “open weight” but not “open source” because we still don’t know the dataset and the training code. LLM360 seem to be fully open source by that definition and releases even the checkpoints!

    Performance wise, a bit lagging (under a Llama2 of the same size) but all the tools are there to improve it!