Understanding FlashAttention Pt 1: Personal Notes

I rebuilt FlashAttention in Triton to understand the performance archaeology

FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttention