Using my tensorflow model in EA

 

Hi ALL,

I spent the last month understanding how to test my tensorflow model in a EA in MQL5.

Doing that in real time, is pretty simple: create a docker container and connect to that via SocketCreate or WebRequest. That is it!

But my problem is how to test in Strategy Tester the model itself. The network function do not work in Tester and what I found as workaround either do not work or it is so complex that I cannot believe it should be the case.

I though about DLL, but creating a DLL from tensorflow, it seems not to be so simple.

Any idea?

I consider this post quite relevant since more and more people are developing their model with tensorflow.

Thanks and cheers,

Marco

 

Hello Marco,

I am trying to do the same thing (using tensorflow model in strategy tester MT5).

Dit you find a way to get this working ?

thanks,

Gino.

 
marco82ger:

Hi ALL,

I spent the last month understanding how to test my tensorflow model in a EA in MQL5.

Doing that in real time, is pretty simple: create a docker container and connect to that via SocketCreate or WebRequest. That is it!

But my problem is how to test in Strategy Tester the model itself. The network function do not work in Tester and what I found as workaround either do not work or it is so complex that I cannot believe it should be the case.

I though about DLL, but creating a DLL from tensorflow, it seems not to be so simple.

Any idea?

I consider this post quite relevant since more and more people are developing their model with tensorflow.

Thanks and cheers,

Marco

Same problem here was there an easy solution?

 
Especially because inside the strategy tester execution times are crucial. Therefore doing Network request-response queries will slow down the whole process.

A workaround can be using named pipes as çonnection model.

So you will need a named pipe server inside your tensor environment. Then connect using named pipes from inside the ea.


Reason: