Machine learning in trading: theory, models, practice and algo-trading - page 2965

[Valeriy Yastremskiy]

Maxim Dmitrievsky #:

I didn't get into his reasoning, it takes him a long time to get his point across.

but Katschyik was right about the big bang (according to the latest observations), about black holes it is still unclear :)

Subtitles, he speaks clearly, you can read)

[mytarmailS]

Valeriy Yastremskiy #:

Subtitles, he speaks clearly, you can read)

Why download subtitles? ))

А. Redozubov in 30-100 years will be something like Einstein now, a genius in a cube...

By the way, GPT4 came out, and nothing special happened, this architecture will not give real AI how many trillions of parameters do not load as he says.

[Vladimir Perervenko]

Renat Fatkhullin #:
Who has tried to convert their models to ONNX and run in the tester?

Could you suggest a Py package to convert MOJO -> ONNX or JSON -> ONNX. Not all models can be converted directly to ONNX (e.g. H2O and some others).

[Rashid Umarov]

The first link in the search gives the result https://docs.h2o.ai/h2o/latest-stable/h2o-docs/productionizing.html

[СанСаныч Фоменко]

mytarmailS #:
About AI

h ttps://youtu.be/AW-4yBTf-XM

Redozubov is a technician, graduated from applied mathematics, and therefore has no special education to talk about AI. A dilettante.

Read Chernigovskaya, at least the man has a profile education, which he uses all his life, is widely recognised in the world.

[mytarmailS]

СанСаныч Фоменко #:

Redozubov is a technician, graduated from applied mathematics, and therefore has no special education to talk about AI. A dilettante.

)))) I'm not even gonna say anything.

[Aleksey Vyazmikin]

Renat Fatkhullin #:
Who has tried to convert their models to ONNX and run them in the tester?

I tried converting, but not running - I wanted to wait for working examples.

And I had a collapse - I replaced the project with an old backup. The most annoying thing was that everything was opened in ME and files there started to change their contents automatically before my eyes - it was very painful to watch it - at least they wrote a warning and offered to save the open version....

[Vladimir Perervenko]

Rashid Umarov #:
The first link in the search yields https://docs.h2o.ai/h2o/latest-stable/h2o-docs/productionizing.html.

Thanks I have tried this package, no luck. Only one GBM model is converted. No good.

[Rashid Umarov]

Vladimir Perervenko #:

Thanks I have tried this package, without success. Only one GBM model is converted. No good.

I haven't looked into it, but it says

Converting MOJOs to ONNX format

To convert a H2O MOJO into the ONNX format, use the onnxmltools python package. Currently, only a subset of H2O MOJOs can be converted to the ONNX format:

• supported algorithms: GBM

• supports multinomial distribution with 3 or more classes (use binomial otherwise)

• does not support poisson, gamma, or tweedie distributions

• does not support models with categorical splits

For examples and further documentation, check the onnxmltools project on github.

[Rorschach]

Vladimir Perervenko #:

Thanks I have tried this package, without success. Only one GBM model is converted. No good.

Have you tried https://github.com/PINTO0309/json2onnx?