search

bambata / ga-bitbot

Automatically exported from code.google.com/p/ga-bitbot
GNU General Public License v3.0
0 stars 0 forks source link

ocl_gts.py not calculating correct scores #24

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
Hello,

    During the last two months I have not succeded to convince ocl_gts.py to calculate correct scores. It shows -10000 in both 'previews' of the top 10 and as a local maximum reached. After I noticed that part, it was not a surprise anymore to me that it never sends any BOBs to the gene server =( Confirmed the behavior on both 64 bit linux machine and 64 bit win machine. I have sent you more detailed description by the mail.

Feel free to ask for any further details that may help and thank you for 
helping us all this time =)

Original issue reported on code.google.com by purge...@gmail.com on 24 Feb 2012 at 3:18

GoogleCodeExporter commented 9 years ago 
open cl components were included in the repo by request because some people 
were interested in working on it. The code on the repo should be considered a 
proof of concept and is currently not in a working state (there are bugs).

Original comment by brian.mo...@gmail.com on 25 Feb 2012 at 2:30

GoogleCodeExporter commented 9 years ago 
I understand =) Please, do modify the ocl_gts.py so it displays the short 
explanation upon startup, so we avoid other people doing the same mistake as I 
did [running ocl_gts.py for months in vain =]

Original comment by purge...@gmail.com on 25 Feb 2012 at 6:06

GoogleCodeExporter commented 9 years ago 
I added the module name of the fitness module / trade engine within the 
gene_def itself and the genetic test sequencer (gts) has been updated to load 
the specified module.

The next steps to reimplementing the potential to run open cl scripts:

1) add detection of an opencl module in gts and reroute the execution to the 
specialized opencl gts implementation.

2) update/upgrade the opencl gts implementation to support being loaded as a 
module.

As the ga-bitbot system is becoming more generalized, the opencl code itself 
will be the responsibility of the user to provide. Any examples provided are 
not meant to be implementations of the existing python fitness modules (such as 
bct) and as such will give different results. 

Also, with the gene database now redefined as a data base library keyed off the 
hash of the gene_def, it's no longer even possible to have both python & opencl 
fitness results stored in the same data base.

Original comment by brian.mo...@gmail.com on 2 Dec 2012 at 8:46