Bringing PyTorch Monarch to AMD GPUs

(pytorch.org)

48 points | by gmays 6 hours ago ago

6 comments

  • mountainriver 5 hours ago

    Love monarch, amazing primitives and feels so much lighter than Ray.

    Another awesome add on their part

  • bicepjai 5 hours ago

    So are we at a point where we can’t train llms for fun at home on less expensive AMD cards ?

    • esseph 4 hours ago

      Who was training LLMs on residential AMD GPUs?

      • MrDrMcCoy 2 hours ago

        I am. Most models that can be run locally have extraordinarily out of date training data for certain frameworks that predate breaking changes and important features, so their docs and API schemas need to be put into LoRAs.

    • amelius 5 hours ago

      can't or can?

  • fefe23 3 hours ago

    benchmqarks?