I'm still running into a memory issue on the latest v3.5.0 release. At full core utilization (128 out of 128) on TACC, I'm getting a numpy.core._exceptions._ArrayMemoryError: Unable to allocate 1.38 MiB for an array with shape (425, 425) and data type float64 error at line 361 of inverse_simple.py. Seems like the prior covariance matrix is blowing up the memory here, even though it only requires a very small allocation. Reducing the number of cores to 64 solves the problem.