burning vs sampling computation speed

[Quresh Latif]

I am just noticing based on progress bars for a couple different models I am running that burnin is occurring much faster than subsequent sampling. I am not sure if I just didn't notice this before or if something changed. For one model in particular, the length of burnin vs subsequent sampling are the same (10000 iterations each), yet burnin appears to be completing in 1-2 hours while subsequent sampling appears to be set to take >1 day.

Wondering if anyone else has had the same experience and/or any insights. Is this to be expected? Am I misinterpreting the progress bar?

[Daniel Turek]

Quresh, thanks for your question.  I'm curious about this, since there shouldn't be any noticeable timing difference between the burnin samples and post-burnin sampling.  Let me clarify one thing, when you see the progress bar, which looks like:

|-------------|-------------|-------------|-------------|
|-------------------------------------------------------|

the top line only shows the *scale* of the entire progress bar.  It does not indicate any sampling has taken place.  The lower line indicates how far along the sampling is (as a fraction of the total length of the top bar).  So for example, sampling is only about to begin when this appears as:

|-------------|-------------|-------------|-------------|
|

and the total sampling is 1/2 complete when it appears as;

|-------------|-------------|-------------|-------------|
|----------------------------

Is there any chance that appearance of the top line (the scale) is being conflated with sampling (specifically, the initial burnin sampling) taking place?

On Wed, Jun 8, 2022 at 4:33 PM Quresh Latif <quresh...@birdconservancy.org> wrote:

I am just noticing based on progress bars for a couple different models I am running that burnin is occurring much faster than subsequent sampling. I am not sure if I just didn't notice this before or if something changed. For one model in particular, the length of burnin vs subsequent sampling are the same (10000 iterations each), yet burnin appears to be completing in 1-2 hours while subsequent sampling appears to be set to take >1 day.

Wondering if anyone else has had the same experience and/or any insights. Is this to be expected? Am I misinterpreting the progress bar?

--
You received this message because you are subscribed to the Google Groups "nimble-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nimble-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nimble-users/2a25f129-7e1f-4576-9889-8b73815b49fbn%40googlegroups.com.

[Quresh Latif]

Yes, I think that is the confusion that I was having. I was interpreting the top line as progress for burn-in and the bottom line as progress for post-burn-in sampling. I thought in previous analyses, the top line was not appearing all at once at the start of sampling, so I thought it represented progress for burn-in, but I must be misremembering. Thanks for clarifying!