The BostonHousing Regression example cannot be compiled

Question

The BostonHousing Regression example cannot be compiled

evo11x opened this issue 3 years ago · comments

The BostonHousing Regression example cannot be compiled with MXNet.Sharp 1.5.2
Can I compile it with the current mxnet source from github ?

Deepak Battini · Answer 1 · Sat Apr 10 2021 09:05:06 GMT+0800 (China Standard Time)

The current MxNet is going under some major upgrade, its been built to make use of numpy and its operators. So NDArray and Symbol API's will be deprecated. At this point, I will be working on the revamp and probably not a good idea to use latest source code for any of your work. You can use the source tagged version and see how it goes: https://github.com/deepakkumar1984/MxNet.Sharp/tree/v1.5.2.0

…

On Sat, Apr 10, 2021 at 9:45 AM evo11x ***@***.***> wrote: The BostonHousing Regression example cannot be compiled with MXNet.Sharp 1.5.2 Can I compile it with the current mxnet source from github ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#42>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABQJAKMD5YERAWMELHEAVL3TH6KATANCNFSM42V4EDRQ> .

-- Regards, Deepak

evo11x · Answer 2 · Sat Apr 10 2021 15:06:39 GMT+0800 (China Standard Time)

Thank you, but that example also does not work with the 1.5.2 from NuGet
https://www.nuget.org/packages/MxNet.Sharp
it's not the same as https://github.com/deepakkumar1984/MxNet.Sharp/tree/v1.5.2.0 ?

MxNet.NN cannot be found

Error CS0234 The type or namespace name 'NN' does not exist in the namespace 'MxNet' (are you missing an assembly reference?)

error CS1503: Argument 1: cannot convert from 'int' to 'MxNet.Symbol'
error CS0117: 'ActivationType' does not contain a definition for 'ReLU'
error CS1061: 'Module' does not contain a definition for 'Add' and no accessible extension method 'Add' accepting a first argument of type 'Module' could be found (are you missing a using directive or an assembly reference?)
error CS0103: The name 'LossType' does not exist in the current context

the Gluon logistic regression example works

evo11x · Answer 3 · Sat Apr 10 2021 17:26:38 GMT+0800 (China Standard Time)

My work with MxNet is experimental so I can use a pre-release if you want me to help you with the issues

Deepak Battini · Answer 4 · Sun Apr 11 2021 11:19:52 GMT+0800 (China Standard Time)

If you can please contribute to this repo will be very helpful. I need to test basic numpy functions, if there is any way you can write test cases for all np.* functions and test it out, I can fix the issues.

…

On Sat, Apr 10, 2021 at 6:56 PM evo11x ***@***.***> wrote: My work is experimental so I can use a pre-release if you want me to help you with the issues — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#42 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABQJAKNUU7CRUXKT4LILY7TTIAKVVANCNFSM42V4EDRQ> .

-- Regards, Deepak

evo11x · Answer 5 · Sun Apr 11 2021 17:41:20 GMT+0800 (China Standard Time)

I am a Windows C++ developer since 1996, with some C# experience about 4 years, but I have zero experience with python/numpy and minimal with github.
I need a C# deep learning api like this which supports CUDA 11 (RTX3060) to implement it in one of my C# projects, for the start I also can use the CPU.
I have tried CNTK but it does not work with CUDA 11 and is a discontinued project, tensorflow is too much python and java oriented so I believe MxNet would be a better choice. I would be glad to contribute with something which I know.

Deepak Battini · Answer 6 · Sun Apr 11 2021 19:18:18 GMT+0800 (China Standard Time)

No worries, give me 1-2 week of timeframe to publish the preview version. Will make sure the examples are running.