Machine learning in trading: theory, models, practice and algo-trading - page 2935

New comment
 
Renat Fatkhullin #:

https://www.mql5.com/ru/docs/python_metatrader5/mt5copyticksfrom_py

Python training is done outside the tester by yourself. Please read the documentation section about Python.

Sorry, the question was in relation to today's average power, 8 cores or 16 threads, 32 gigs of memory average TC algorithm (well 1000 calculations per tick for example) is possible in an hour a couple of years.

 
Aleksey Nikolayev #:

I took the model with the lowest version from here, put it in a common folder, ran the script:

No description of error 5801 found. My system: Windows 10 build 19045. According to the table should work.


I guess I need to install something, but I don't have a clear understanding of what it is yet.

 
Aleksey Nikolayev #:

Probably need to install something, but no clear idea what it is yet.

I've had 5801 and 5802 crash, I guess we should wait for examples
 
Aleksey Nikolayev #:

Probably need to install something, but no clear idea what yet.

We need to wait for examples at least. The promised functionality is encouraging.

Only :" ONNX Runtume will be rewritten and redesigned not to depend on obsolete system libraries" .

Waiting

 
Vladimir Perervenko #:

We'll have to wait for at least some examples. The promised functionality is encouraging.

Very encouraging.

Vladimir Perervenko #:

Only :" ONNX Runtume will be rewritten and redesigned to not depend on obsolete system libraries".

Perhaps this is to avoid drowning in the zoo of ONNX versions. Or, perhaps, to avoid mandatory installation of Visual Studio - nowadays, it seems, there is no way without it.

Maxim Dmitrievsky #:
I had 5801 and 5802 crashing, I guess we should wait for examples

Waiting

 
Aleksey Nikolayev #:

Probably need to install something, but no clear idea what yet.

Well, it says: ONNX version = 1.2.1; opset version = 7. It is not clear what it is, but obviously your settings do not meet these requirements.

 
Renat Fatkhullin #:

For Microsoft's in-house ONNX Runtime, the minimum version claimed is Windows 10 build 1809.

And in C:Windows\System32\onnxruntime.dll there is a very old version 1.10. At the moment the desirable version is 1.14.

We will write step-by-step articles on how to use ONNX functions.

Good day. When approximately to wait for articles?

 

New bugs:

   MQL_DEFINE_ERROR(ERR_ONNX_INTERNAL                       ,5800)
   MQL_DEFINE_ERROR(ERR_ONNX_NOT_INITIALIZED                ,5801)   // ошибка инициализации ONNX runtime API
   MQL_DEFINE_ERROR(ERR_ONNX_NOT_SUPPORTED                  ,5802)   // фича или значение неподдерживается нашей реализацией ONNX
   MQL_DEFINE_ERROR(ERR_ONNX_RUN_FAILED                     ,5803)   // ошибка запуска ONNX runtime API
   MQL_DEFINE_ERROR(ERR_ONNX_INVALID_PARAMETERS_COUNT       ,5804)   // передано неверное количество параметров в OnnRun
   MQL_DEFINE_ERROR(ERR_ONNX_INVALID_PARAMETER              ,5805)   // некорректное значение параметра
   MQL_DEFINE_ERROR(ERR_ONNX_INVALID_PARAMETER_TYPE         ,5806)   // некорректный тип параметра
   MQL_DEFINE_ERROR(ERR_ONNX_INVALID_PARAMETER_SIZE         ,5807)   // некорректный размер параметра
   MQL_DEFINE_ERROR(ERR_ONNX_WRONG_DIMENSION                ,5808)   // не задана размерность тензора
 
For an example, you can see the public project ONNX.Price.Prediction in the section of MetaEditor common projects.
 
Vladimir Perervenko #:

Well, it says: ONNX version = 1.2.1; opset version = 7. It is not clear what it is, but obviously your settings do not meet these requirements.

My system: Windows 10 build 19045. According to thetable it should work (1.2.2 > 1.2.1 and 19045 > 19041).

Based on the second line of the table: Windows 10, version 2004 (build 19041) ONNX version = 1.2.2, 1.3, and 1.4 opset version = 7, 8, and 9.

PS. Perhaps the problem is that I have the Home version of Windows.
1...292829292930293129322933293429352936293729382939294029412942...3590
New comment