Split Q,K,V vs Fused QKV Projections in diffusers Discussion

`diffusers` generally splits fused QKV projections to separate Q, K, V projections in its checkpoint format (for example, for diffusion transformer models). I am interested in collecting feedback on this design choice and any problems which may arise from it. To be honest upfront, I think it is unlikely that we will change this design in the near future, but we would appreciate your feedback nevertheless.

EDIT: the reasoning for using split Q, K, V projections is described in more detail in https://github.com/huggingface/diffusers/issues/14003#issuecomment-4774944332.

Comienza con el cuerpo del issue y el comentario enlazado, que explican la justificación del diseño de checkpoints con Q, K, V separados frente a un QKV fusionado. No se identifican archivos, pruebas, puntos de entrada, objetivo de implementación ni criterios de finalización; no se define qué significa estar terminado más allá de recopilar comentarios sobre el diseño.

Escrito por el modelo de indexación a partir del texto del issue.