Lc0 stops calculating too soon

[Robaloooo]

Lc0 is only thinking for 15 seconds when set to 30 seconds.
This is with top 4 moves selected.

[Robaloooo]

This is still happening. Sometimes only calculates 10 seconds instead of 30.

Is it possible that when Lc0 was upgraded to 0.28 it reverted back to the default setting to think only until it knows the top move will not change?

[Robaloooo]

I am reminded of this thread:

LCZero doesn't search for 30 seconds

At first I thought it was just my imagination, but according to the “time” numbers in the logs LCZero is not searching for 30 seconds. The average search lasts only about 18 seconds. 16 seconds was the shortest I observed and the longest was 22. Weights files are not a factor, it happens with the smallest all the way up to the biggest.

[chendry]

Doh! You’re indeed correct. This is a bug, and it was introduced in late August. Essentially the input filter for the engine which dynamically sets SmartPruningFactor whenever it sees MultiPV getting set to anything > 1 was not running.

I just deployed a fix – everything should be working as intended now. Thanks for reporting this!

[puttutathy]

lc0 not thinking 30 sec some time only 10 second and plying week no profit for gpu

[chendry]

It sounds like you may have MultiPV set to 1. In that case, NCM keeps SmartPruningFactor at it’s default value, so Lc0 will abort the search early. From the documentation:

Do not spend time on the moves which cannot become bestmove given the remaining time to search. When no other move can overtake the current best, the search stops, saving the time. Values greater than 1 stop less promising moves from being considered even earlier. Values less than 1 causes hopeless moves to still have some attention. When set to 0, smart pruning is deactivated.

[puttutathy]

@chendry

MultiPV set to 1

yes MultiPV set to 1 for strong move

[chendry]

So I think everything is working correctly then? The search may (and probably will) stop before the 30 seconds, but that’s expected.