Errors, bugs, questions - page 1993

 
And occasionally some of the agents drop out on 7 and don't restart themselves. What to do? Maybe it's the wrong system? Or may be something else needs to be adjusted? No antiviruses or anything else on the machine
 
Anton Ohmat:

Good afternoon.

Question. I have set up a local network at home with 256 agents on xeons. Metatrader does not want to use 30% of the agents for normal testing. And in genetic algorithm about 60 cores are idle.

Suggest where the problem might be.

Maybe in the metatrader, what settings have to be done?

Perhaps, the computer you are testing on has no time to communicate with agents via home network? Try to leave only one agent on this computer and disconnect the rest. Also just before genetics and optimization recompile your Expert Advisor.

 
Anton Ohmat:

Good afternoon.

Question. I have set up a local network at home with 256 agents on xeons. Metatrader does not want to use 30% of the agents for normal testing. And during genetic algorithm around 60 cores are idle.

Suggest where the problem might be.

Maybe in the metatrader some settings need to be done?

Optimization involves analysis and consolidation of data on the control terminal, so splitting up into too small batches of jobs is not optimal - there will be a lot of overhead. If all the capacity is not taken up it means that the task is not complex enough, i.e. it is more efficient to solve it with a smaller number of agents. In fact, this issue has already been raised - search the forum.

 
Anton Ohmat:

Question. I have set up a local network at home with 256 agents on xeons. Metatrader does not want to use 30% of the agents in normal testing. And with genetic algorithm about 60 cores are idle.

Only on genetics?

Do each of the following

  • Shut down all the Agents which were involved in Optimization, and leave only the "problematic" 60 in Ready state. Will they be used in Optimisation afterwards?
  • Look at the number of Optimization passes. Try to increase their number, as with a small number it is impractical (especially with GA) to use a large number of Agents.

 

I don't know if this is a bug or not. When you look at the line graph after optimisation.

If the values are scaled down, nothing is displayed

 

You can fix it on MQL5 without the possibility of fixing it on the same MQL5 in the following way.

Create templates through ChartSaveTemplate, until the disk is full. And to delete them... no way!

 

Question to developers.

It turns out that out of 256 agents, half of them are idle on genetic optimization.

Is it possible to test 2 different strategies simultaneously in 2 programs by adding the same agents to both terminals?

 
Anton Ohmat:

It turns out I have 256 agents and half of them are idle on genetic optimisation.

Is it possible to test 2 different strategies simultaneously in 2 programs by adding the same agents to both terminals?

If it wasn't, the Cloud wouldn't work. Feel free to run the required number of terminals and go ahead!

 
fxsaber:

If it wasn't, the Cloud wouldn't work. Feel free to run the right number of terminals and go for it!

Thank you
 
fxsaber:

If it wasn't, the Cloud wouldn't work. Feel free to run the right number of terminals and go for it!

It doesn't work in this mode. Knocks out agents as failed and that's it. Bummer. What's the difference between cloud - specifically limited?
Reason: