Arnaud A

Very limited proof of concept PR to wrap around LAPACK's `dpstrf` and carry out the Cholesky Factorization of PSD matrices. In its current state of `ndarray_linalg` does not handle this...

I think I am missing something when going over the Triton implementation of LayerNorm in the tutorial. In the forward, we compute the mean as: ```python _mean = tl.zeros([BLOCK_SIZE], dtype=tl.float32)...