https://github.com/ELS-RD/kernl/blob/91e2cd92db44d503874d39a9f6dec42c9f481a8e/src/kernl/implementations/layer_norm.py#L177C13-L177C63 Why not `block_delta = tl.sum((a - block_mean) * (a - block_mean), axis=0)` ?
https://github.com/ELS-RD/kernl/blob/91e2cd92db44d503874d39a9f6dec42c9f481a8e/src/kernl/implementations/layer_norm.py#L177C13-L177C63
Why not
block_delta = tl.sum((a - block_mean) * (a - block_mean), axis=0)?