Logging training results with WFO & NumCores

When I run training with WFO cycles and NumCores=-1 I notice there only appears to be 1 core writing to the _train.log file.

I have an 4 core processer. So when running NumCores=-1 I get 7 parallel threads running zorro instances. The _train.log files looks like the below (showing only lines with "walk").

Is there any way to report all training when running parallel training?

Only with a normal training run. The extra cores don't write a log.

That's unfortunate... I have a Ryzen 3950x on the way.

On another note. Is there any way to log more than the return from the objective() function in the _train.log files? Such as other trade statistics? I have considered more complex selection of optimized values with a filtering approach documented by myersanalytics.com. This would have to include a parsing and post process effort to write the .fac files.

To get the trade statistics I guess perhaps I could trying writing to a file manually inside the objective function. Maybe this would allow logging anything required and take care of the logging with extra cores. But maybe file locks would be an issue. Not too worried about order if I can get the WFOCycle into the log file.

Thanks for the quick response!

I believe this behavior can be overridden by setting LogNumber, such as LogNumber = Core. Support told me this feature was added as of version 2.23.6.

I'm not sure if this was documented in the manual though.

Excellent. I have not tried testing on my above ideas yet, I will try this first. Thanks for the tip!

I see this in the documentation:

LogNumber
When nonzero, appends the given number to the log and image files generated by the simulation. By setting LogNumber = TotalCycle, different log files or chart images can be generated for any cycle. Set this variable before calling any function that writes something into the log.