problems with initial model

[doloacu]

Hi, I'm suggesting a NEXUS tree and the ML looks really nice, but I cannot make BEAST works. This is how the error message starts:

Creating the MCMC chain:
chain length = 200000000
operator adaption = true
adaptation delayed for 2000000 steps
Underflow calculating likelihood. Attempting a rescaling... (patterns)
Aug 15, 2025 11:48:21 AM dr.app.beast.BeastMain
SEVERE: Error running file: rsv-b-desafio-v12-mafft-para-beauti-1.xml
Aug 15, 2025 11:48:21 AM dr.app.beast.BeastMain
SEVERE: The initial model is invalid because state has a zero probability.

If the log likelihood of the tree is -Inf, this may be because the
initial, random tree is so large that it has an extremely bad
likelihood which is being rounded to zero.

Alternatively, it may be that the product of starting mutation rate
and tree height is extremely small or extremely large.

Finally, it may be that the initial state is incompatible with
one or more 'hard' constraints (on monophyly or bounds on parameter
values. This will result in Priors with zero probability.