add ability to turn on normformer settings, given @borisdayma reporte…

…d good results and some personal anecdata

makes more sense to keep this as True as default, for stability

provide option to l2norm the output of the diffusion prior

fix issue with mixed precision and gradient clipping

fix calculation of adaptive weight for vit-vqgan, thanks to @CiaoHe

switch to using linear attention for the sparse attention layers with…

…in unet, given success in GAN projects

more attention

project management

let researchers do the hyperparameter search

add MLP based time conditioning to all convnexts, in addition to cros…

…s attention. also add an initial convolution, given convnext first depthwise conv