ValueError: conflicting sizes for dimension 'chain': length 1 on the data but length 2 on coordinate 'chain'

I removed both the tags. But the question is now, how I will control the sample size from the trace in posterior prediction. And similarly how we will do it when have more chains.