save_interval is not respected in kima.run

[Kamuish]

Hi,

I have noticed that when setting the

kima.run(
model,
steps=n_steps,
num_threads=30,
new_level_interval=4000,
save_interval=500,
)

save_interval=500, kima still prints " Saving particle to disk. N = " every 50 samples. The only way to change the interval between disk "saves" seems to be to change the "print_thin" flag. I am not sure if this is an issue only with the terminal output, or with the frequency of I/O operations

[ThomasBaycroft]

Hi,

A bit late on this, but I think this is a confusion with the word steps being used in two different ways. I interpret it as:

• save_interval is the number of iterations the sampler goes through before saving/recording a sample
• Each of these "saves" is a step in terms of steps=n_steps,
• Then every 50 saves, the output prints to keep track of the run, this can be reduced with "print_thin" but this shouldn't affect the actual sampling process in any way

Maybe "n_steps" should be renamed "n_saves"?

[j-faria]

Indeed, the argument names are a bit confusing. But for the sake of backwards compatibility (and consistency with DNest4), I would prefer to not change them. I will add some documentation about this.