Roffild's library - page 9

New comment
 
Roffild:

If the Python path on the Agents computer is the same as the home computer and the PATH variable also contains the path, it should work...

Where is this PATH variable located?

 
Fixed tester crash on build 2025 due to PythonDLL.
 
Waiting for bug reports...
 

The execution status in PythonDLL is now checked by code:

if globals().get('__PythonDLL__'):
    print('run in MetaTrader')
elif __name__ == '__main__':
    print('run as script')
 

Switching to plain GIL because many popular Python libraries are not compatible with the subinterpreter.

You can build with PYTHONDLL_SUBINTERPRETERS if needed.

MetaTrader 5 (MQL5) + Python 3 DLL для Forex, CFD и Futures
  • roffild.com
Использование MetaTrader с Python 3 на финансовых фондовых биржах, Forex, CFD и Futures. Из MetaTrader можно получать котировки в Python, но нет полноценной связи между ними. Пост одного из разработчиков. Главная идея и отличие этой обертки от остальных: обмен данными между MQL и Python через заранее созданные функции. Это самый быстрый и...
 

Updating the PythonDLL again

Some use sys.stderr.write(), but with no console, sys.stderr=None. We had to make a stub.

 

Roffild, good afternoon.

Do you have a library available for python, version 3.6? I use TensorFlow and its latest version only supports 3.6.

If available, please provide a link to the file.

Thank you.

 
canonier:

Roffild, good afternoon.

Do you have a library available for python, version 3.6? I use TensorFlow and its latest version only supports 3.6.

If available, please provide a link to the file.

Thank you.

TensorFlow 1.13 and 1.14 work fine with Python 3.7 myself. No problems with Pytorch either.

 
Roffild:

TensorFlow 1.13 and 1.14 work fine with Python 3.7 myself. No problems with Pytorch either.

They do not work if the processor does not support some processor instructions (I thinkAVX isnot present inall processors ), wrote https://www.mql5.com/ru/forum/307970/page5#comment_11086500

I wrote about it and it is interesting that on my laptop does not work with TensorFlowSharp under c# , but on PC no problems, googled the info need to recompile tensorflow.dll itself for c++, it is in the TensorFlowSharp package


In general, not all hardware can run the same version of TensorFlow , for my laptop only found tensorflow-1.6.0-cp36-cp36m-win_amd64.whl on GitHub it installed, but as written above for Python 3.6 only


here's the GitHub to find what I'm writing about:https://github.com/fo40225/tensorflow-windows-wheel

 

The message"Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2" appears on newer CPUs as well. This is a known bug in TensorFlow. It does not affect the Python version in any way.

Since TensorFlow 1.13 in Anaconda is built with Python 3.7

Pytorch 1.0 is also built under Python 3.7

I can build a library under Python 3.6, but I've been using Python 3.7 since January and have no problems.

1234567891011
New comment