advector error inside the
complex eigenvector normalization path. This appears to be platform and
toolchain sensitive: Apple silicon, especially M1 (but not M3), does not
provide the same extended precision behavior as some x86 paths, and the
reduced floating-point headroom can expose fragile complex eigen/AD
calculations. Added a lazy runtime capability check for RTMB AD eigen
support. Normal platforms continue to use RTMB’s AD eigen constraints.
If the capability check fails, the package falls back only for
eigenvalue constraints: constraint values are computed with numeric
base::eigen() and constraint Jacobians are computed by
central finite differences. The likelihood, scores, and other TMB/RTMB
AD paths remain unchanged.tsissm_capabilities() to report the detected
RTMB AD eigen capability, source of the decision, RTMB/TMB versions, and
platform details.tsissm.rtmb_ad_eigen option under
estimate() so users can force RTMB AD eigen constraints or
force the finite-difference fallback for diagnostics.as_flextable() by printing the model summary directly.
Need a high-speed mirror for your open-source project?
Contact our mirror admin team at info@clientvps.com.
This archive is provided as a free public service to the community.
Proudly supported by infrastructure from VPSPulse , RxServers , BuyNumber , UnitVPS , OffshoreName and secure payment technology by ArionPay.