Discussion of article "Neural networks made easy (Part 36): Relational Reinforcement Learning" - page 2

New comment
 
DENKZ strategy tester? If I understand correctly, in the tester, again, a virtual sandbox is created, and even if the file is forcibly copied to this directory of the tester, then when the tester is restarted, all files from it are deleted. Here is this directory as an example. C:\Users\ACER\AppData\Roaming\MetaQuotes\Tester\36A64B8C79A6163D85E6173B54096685\Agent-127.0.0.1-3000\MQL5\Files The network is naturally built on new and untrained. Or how can I solve this problem otherwise?

Good day,

You can use the common directory. Tester agents work with it too. To do this, you need to change the flag when loading the model.

 

Tôi đã bật OpenGL nhưng tôi gặp lỗi như hình bên dưới. Làm thế nào tôi có thể chỉnh sửa nó?

Files:
Error001.PNG  32 kb
 

Hi, everybody. I have a problem:

20 part ae , ae2 goes and starts.

21 part vae, vae2 compilation error 'pow' - ambiguous call to overloaded function NeuroNet.mqh 4467 38

22 part compiles but at startup right after loading OpenGL driver the screen blinks and stops.

23 part AE starts

rnn_vae compiles compiled but at startup immediately after loading OpenGL driver the screen blinks and stops.

vae generates compilation error

36 reinforce zero code error

rrl learning compiles, but at startup immediately after loading the OpenGL driver the screen blinks and stops


OpenGL

 

Hi Dmitry Gizlyk,

Thanks for your wonderful articles.

Please help! When I try to train in Strategy Tester using Ryzen 9 6900hx (APU), I got this error and the EA had no transaction.

How to fix this problem bro?

 

Hi Dmitry,
Thank you for the awesome work! This is the best tutorial so far on the net regarding ML on MQL5 platform
Details are well explained and can be understood by new learner.

Following your tutorial, I've run Strategy Tester and apparently, only one of my processors were used despite there are 12 available as shown in the picture below

Is there any way to activate all cores instead of just one?

OS Windows 11 build 22H2
OpenCL Support 3.0
CPU Intel i5-12400 ghz.html)
GPU Intel UHD Graphic 730 (integrated)
RAM 16gb

OpenCL are already enabled in Metatrader's settings.


Thanks for the detailed tutorial!

 
sy4rul #:

Hi Dmitry,
Thank you for the awesome work! This is the best tutorial so far on the net regarding ML on MQL5 platform
Details are well explained and can be understood by new learner.

Following your tutorial, I've run Strategy Tester and apparently, only one of my processors were used despite there are 12 available as shown in the picture below

Is there any way to activate all cores instead of just one?

OS Windows 11 build 22H2
OpenCL Support 3.0
CPU Intel i5-12400 ghz.html)
GPU Intel UHD Graphic 730 (integrated)
RAM 16gb

OpenCL are already enabled in Metatrader's settings.


Thanks for the detailed tutorial!

Hi, at Strategy tester you can see only one core use. It used by mql programme, not OpenCL. OpenCL use GPU or CPU cores in system outside Strategy tester monitor. There are several ways to see the resource consumption of an OpenCL programme in Windows:

1. Use performance monitoring software such as MSI Afterburner or GPU-Z, which display GPU usage and other system components. They can also show what portion of resources each OpenCL program is using.

2. Use profilers such as AMD CodeXL or NVIDIA Nsight Visual Studio Edition. They allow you to analyse an OpenCL program and display which parts of the code consume the most time and resources.

3. Use the OpenCL API to gather statistics. This allows you to programmatically obtain information about the use of OpenCL resources, such as memory usage or core performance. You can use the Performance Counters for OpenCL (PCPerfCL) library to gather this information in Windows.

4. Use profiling tools such as Intel VTune Amplifier, which can help you see how a program uses processor and other system component resources.

 

Good afternoon Dimitri!

Please help me to run your Expert Advisor from this article. I have tried everything to make it work, but alas it does not work properly.

The problem is as follows: the Expert Advisor starts in the tester and begins the test in the tester normally, counting down the seconds, green bar, no errors in the log, normally sees the video card and selects it. But there is 0% load on the video card. It's like it's not counting anything on it. In Common\Files there are 2 files with extensions icm and nnw with size 1 kb. When I try to restart the test in the tester again, it warns that it cannot initialise and the test does not start. If you restart MT5 and delete the files created by this EA in Common\Files, it starts normally, but also does not use the video card and creates again these files of 1 kb. and so on and so forth.

I tried to take NeuroNet.mqh files from the next article (that you posted there in the comments) and replace the one in the article with it - it didn't help. I tried selecting a small time period in the tester (1 month, 1 week, 2 months, etc.) also didn't help.

How to start it? The Expert Advisors from previous articles run normally and use the video card correctly.

There is also a problem with the Expert Advisor from the next article 37, 38. On the contrary, there is no progress in the tester, but the video card is used to the maximum and so at least 5 hours, even 10 hours.

The Expert Advisor from article 39 worked fine. There I chose the history of more than 1 month and it did not create the database, but I chose 1 month and it created the database normally. The rest of its parts worked normally.

Log of work in the tester
 

This is a real gem, Dmitriy!

I am a data scientist with hands-on developing various machine learning algorithms on time-series data, and yet I find your series informative and well-written. This is a valuable basis to build on. 

Thank you for sharing your work.

Bijan

12
New comment