A new Zorro version is available:

http://opserver.de/down/Zorro_188.exe

This version will be uploaded to the Download page when no bugs are found in the next time. Zorro now supports order flow analysis, COT report analysis, more price sources and cryptocurrency exchanges, and many more new features and improvements. The full list of new features and bugfixes can be found on http://manual.zorro-project.com/new.htm. Please note that the trade format has changed, so you cannot continue old trades with the new software.

Please test everything and report any problems here!

The training becomed very slow after the first walk for the core. I had the same problem with the previous beta version.

So the second walk trains slower than the first walk? Does this only happen when you use cores? Only with your script, or also with the included scripts?

Is the Bitfinex plugin using the Websocket API to get price data, and if so, can we use the callback function to get price updates? I'm thinking it would be interesting to attempt to implement an HFT system for a crypto exchange :-)

I tested Workshop5 and Workshop6 and looks they already started running slow for all cores. I was able to get smaller script that have the problem. It do not only happens with multi core. On the script below with 2 cores the walk 1 and walk 2 runned fast. The walk 3 and walk 4 runned slow. The same happened after disabling multi core. I increased the number of cores and the first walk for all cores runned fast and it did make the walk train fast even for 2 cores and for no multi core. The speed up was preserved after a Zorro restart.
Commenting out "while(asset(loop("EUR/USD")))" while keeping "EUR/USD" selected did make this code runs slows. However with my script half cores started running fast and half slow.

The Bitfinex plugin uses the REST API, not the Websocket API.

I think I know what you mean with the slow training. Once historical data is loaded in the cache, the script runs much faster. This happens at the begin of training because the first part of the data is already in the cache. If you train again, it should be faster - try it. But there is room for improvement and we'll look into it.

I did some tests. When it becomes slow each parameter step take more time to run. After the walk is finished all steps for such walk run faster. I added a timer at the begin and the end of the run function on the script above and it needs more time to complete. If I let Zorro finish the training all walks run faster for such script. However I see no difference for other scripts. I tested on wine.

Try again on Windows and see if you get the same behavior.

You are nobody to say what I should do.

My Contracts per lot has changed to 1, how do I take it up 1000?