How much training consumption is required for visual validation?

[Aziily]

Hello, I am trying to train a pyramid flow model from scratch. Due to limited training resources, I want to first check the validity of my training code. I think the most intuitive way is to see whether the model can generate something corresponded with the prompt.

But I cannot afford several full experiments, so I wonder at least how much Batch Size x Steps is required for the generation in accordance of the prompt?

I am just trying the T2I phase at present. I will appreciate it if you can offer data for both T2I and T2V phases. Thank you again.