Closed decoda-huanyi closed 1 year ago
@decoda-huanyi Hey I'm still working on the actual model architecture, i need help with implementing it.
Would you like to help integrate it
I am acturally interested. Which dataset shall we start with? In my opinion, this dataset should have very long samples so that we can demostrate the power of longnet. I am not sure. Could we just try it on a noval or, you know, some other modals like music or images?
@xiehuanyi Hey we're about to start training on the enwiki dataset there's a competition for some money on whoever can train the best model on 1gb of wikipedia data!
Where can I find some examples showing how to train longnet on a real dataset?
Upvote & Fund