I am running into some performance issues when specifying and sampling from a linear model with b-splines in PYMC (chapter 4, R Code reference 4.76). Everything works fast as long as I keep the number of knots and/or the degree low enough (below 10 knots and 2 degrees). As soon as I increase one of the parameters above the mentioned values, the model no longer samples within ~30 seconds, but takes 20 minutes+.
Has anybody else run into similar issues? I am currently puzzled by the rapid performance drop...
Any help or pointers would be appreciated.
Hello.
I am running into some performance issues when specifying and sampling from a linear model with b-splines in PYMC (chapter 4, R Code reference 4.76). Everything works fast as long as I keep the number of knots and/or the degree low enough (below 10 knots and 2 degrees). As soon as I increase one of the parameters above the mentioned values, the model no longer samples within ~30 seconds, but takes 20 minutes+.
Has anybody else run into similar issues? I am currently puzzled by the rapid performance drop... Any help or pointers would be appreciated.
Code
Versions: name : pymc version : 5.9.1
name : numpy version : 1.25.2
Best htw