Will ONNX models made in R run in linux? - General

Aleksey Nikolayev 2023.03.01 10:54 #29381

Vladimir Perervenko #:

I totally agree! It's a lot of work. Two questions:

1. Can models created using GPU be converted to ONNX?

2. To get the model prediction in ONNX in the MQL5 Expert Advisor, we don't need Python, do we? The calculations will be performed by the onnxruntime environment?

In theory, this is probably so. But it may turn out in practice that only models trained in Python are suitable, and only in some unique environment of packages, utilities and compilers (taking into account their version numbers). At least that's how I understood Renate's hint.

From theory to practice Optimal strategy under statistical MetaTrader 5 Python User

Renat Fatkhullin 2023.03.01 10:59 #29382

Vladimir Perervenko #:

I totally agree! It's a lot of work. Two questions:

1. Can models created using GPU be converted to ONNX?

2. To get the model prediction in ONNX in the MQL5 Expert Advisor, we don't need Python, do we? The calculations will be performed by the onnxruntime environment?

1. Yes, it depends on the source engine, which can convert to ONNX within the existing opset

2. You don't need Python to execute the model, all you need is one EX5 file and a terminal. Even ONNX Runtime libraries won't be needed any time soon

Learning ONNX for trading How to Start with Something Interesting in Financial

Aleksey Vyazmikin 2023.03.01 11:02 #29383

Renat Fatkhullin #:

2. You don't need Python to execute the model, all you need is one EX5 file and a terminal. Even ONNX Runtime libraries will soon be unnecessary

And then will this thing work on semiconductor?

Vladimir Perervenko 2023.03.01 11:36 #29384

Renat Fatkhullin #:

1. Yes, it depends on the source engine that can convert to ONNX within the existing opset

2. You don't need Python to execute the model, all you need is one EX5 file and a terminal. Even ONNX Runtime libraries won't be needed any time soon

1. and we have opset=14 ? I use PyTorch(1.12.1) and TF(2.10).

Предупреждение: TensorFlow 2.10 был последним выпуском TensorFlow, который поддерживал GPU в родной Windows. 
Начиная с TensorFlow 2.11 , вам нужно будет установить TensorFlow в WSL2 или установить tensorflow-cpu и, 
при желании, попробовать TensorFlow-DirectML-Plugin.

2.Extremely important and this will be the biggest advantage of this integration. Great

Aleksey Nikolayev 2023.03.01 11:47 #29385

The most interesting thing is whether ONNX models made in R will run.

A list of supported ONNX versions and opsets would be useful.

Aleksey Nikolayev 2023.03.01 12:00 #29386

Aleksey Nikolayev #:

The most interesting thing is whether ONNX models made in R will run.

We can even aggravate the question. Will ONNX modelsmade in R run in linux?

mytarmailS 2023.03.01 12:02 #29387

Aleksey Nikolayev #:

The most interesting thing is whether ONNX models made in R will run.

A list of supported ONNX versions and opsets would not be unreasonable.

I support the question

Vladimir Perervenko 2023.03.01 12:10 #29388

mytarmailS #:
Seconded question

What R packages do you use to create models?

Evgeny Dyuka 2023.03.01 12:43 #29389

Aleksey Nikolayev #:

I'm posting the project in the archive just in case.

Thank you, but I'm jumping off ... it's too expensive to go any further, I don't have that much time ))

mytarmailS 2023.03.01 12:47 #29390

Vladimir Perervenko #:

What R packages do you use to create models?

The usual ones, the same "random forest" "intrees" "arules" "arules sequence" "dbscan".

But that's 5%.

And the other 95% is self-written stuff.

If you can't embed any of your own code in onnx, then I'm not interested in it.

Also half of my algorithms are rebooted at every step, as I've heard this doesn't work with onnx.

Open opposite order Can someone explain me strange profit-loss calculation -

Machine learning in trading: theory, models, practice and algo-trading - page 2939