Cholesky decomposition performed for every sample(?)

[junpenglao]

Discussed in #758

^{Originally posted by jonny-so December 4, 2024}
Hi, unless I have missed something, it looks as though (e.g.) the NUTS sampler implementation creates a new Euclidean metric object for every sample

blackjax/blackjax/mcmc/nuts.py

Line 123 in 65ae00e

metric = metrics.default_metric(inverse_mass_matrix)

which in turn performs a Cholesky decomposition of the (inverse) mass matrix. Is that correct? It should only be necessary to do this at the end of the adaptation windows. In principle the compiler could be eliding the redundant calls, but I would guess that is asking too much of it in this case.

[AdrienCorenflos]

c.f. my answer, I'd like for use to confirm there is a problem/where the problem is before designing a fix.