Computed two more interesting posteriors. Computing the posterior for mcycle_gp-accel_gp turned out to be quite a struggle and I had to push adapt_delta to 0.999 (even though we had discussions of using max 0.99) and max_treedepth to 15. Also I think it is worth noting that I had to "hack" the accel_gp model a bit; I changed a few parameters vector type (of length 1) in parameters block to float to band casted them to back vector type in transformed parameters block which was required by the gpa function in model block. This had be done to handle due to R's conflation of scalars and 1-dimensional arrays. Related issue.
Computed two more interesting posteriors. Computing the posterior for mcycle_gp-accel_gp turned out to be quite a struggle and I had to push adapt_delta to 0.999 (even though we had discussions of using max 0.99) and max_treedepth to 15. Also I think it is worth noting that I had to "hack" the accel_gp model a bit; I changed a few parameters vector type (of length 1) in
parameters
block to float to band casted them to back vector type intransformed parameters
block which was required by thegpa
function in model block. This had be done to handle due to R's conflation of scalars and 1-dimensional arrays. Related issue.