- Lightning Attention
- the lightning attention was introduced in the minimax paper
- in the lightning attention forward pass we can see that there is a mat mul part and we can bind the O_intra with a softmax
- so we have to make a mat mul kernel and softmax kernel & then we will make a lightning attention function to write the forward pass
- i will be going for the naive mat mul kernel & naive softmax kernel
- i will be using my previous implementations for both
