Accessing generation number from inside evaluate()

[andresC98]

Is there any way we could actually access the Generation number (integer) from inside Evaluate() function?

[mikefenton]

Hi Andres,

You can just import the stats.stats object, we track the generation number there.

[andresC98]

Thanks! Using stat.stats['gen'] was just what I was looking for!

[andresC98]

Actually I noticed this does not work if using --multicore option (keeps having stats['gen''] = 0).

Is there any way to workaround this?

[mikefenton]

Hmm, if that's true then it's a bug for sure. If you check the code for algorithm.search_loop.search_loop() you can see it's setting the stats['gen'] parameter in each iteration, independent of the multicore operation. Will investigate further.

[dvpfagan]

Does multi core not make a stats for each pool and then merge after all the runs are done. Might we not be setting the gem in these copies? Dave

…

On Tue 23 Jul 2019 at 09:08, Michael Fenton ***@***.***> wrote: Hmm, if that's true then it's a bug for sure. If you check the code for algorithm.search_loop.search_loop() you can see it's setting the stats['gen'] parameter in each iteration, independent of the multicore operation. Will investigate further. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#99?email_source=notifications&email_token=AAHTHOX6TQ2NJSYQG6L45KDQA24APA5CNFSM4ICFRBYKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2SJEZA#issuecomment-514101860>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAHTHOQPKVNU26ZODBVMDQDQA24APANCNFSM4ICFRBYA> .

[mikefenton]

@andresC98 can you provide the parameters file you used to produce this error? Running tests here and everything appears to be in order.

[andresC98]

Sure! I am using a very simple parameters file, with my custom grammar and fitness function:

CODON_SIZE: 100000
GENERATIONS: 10
MAX_GENOME_LENGTH: 500
GRAMMAR_FILE: mlpGrammar.pybnf
POPULATION_SIZE: 15
FITNESS_FUNCTION: NescNNclasGE

and when running the script I add the --multicore option.

I am using a custom fitness function for classification, following the template (I correctly import from fitness.base_ff_classes.base_ff import base_ff and initialize class as super.__init__() , etc).

As you mentioned, I also import stats.stats and when I am accessing the stats dictionary key for 'gen' inside the evaluation function, but always prints "0". However, running the script (with the exact same code for the fitness function) without --multicore works just fine printing correctly the generation number value.

[MarkusEvo]

Hey, did this Issue ever get solved?

[dvpfagan]

The import of stat.stats works as mentioned in the thread but no fix for the multi core issue has been released. Dave

…

On Thu, 24 Dec 2020 at 13:25, MarkusEvo ***@***.***> wrote: Hey, did this Issue ever get solved? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#99 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHTHOTDADIUT3AQSSECAODSWM6K3ANCNFSM4ICFRBYA> .