Great Work, Can you provide the Training Time and Total Steps of the flux-kontext to align the pretrained result?