No, not Adam, I was referring to the model from the diffusion beats Gans paper.

I never trained such model, just read it. But yeah it's most likely what you said (a.k.a data parallelism)