I pretty much just followed this Tensor Flow tutorial, just used my own data.I’ve tried different models, different numbers of inputs, etc and the picture is more or less the same – constant offset between labels and predictions…