07:51 <Guest7823> Hi Team i have a mlpack model file , now i want to load this file to python for inferencing .

07:52 <Guest7823> could you pls help me on that

07:56 <Guest786> Hi Team i have a mlpack model, how can i load to python for inferencing .

07:56 <Guest786> could you pls help me on that

12:29 <EshaanAgarwal[m]> <zoq[m]> "Let’s move the meeting to..." <- Hi zoq ! Would it be possible to move this meeting later to Thursday or Friday ? I am actually travelling 😬

12:30 <EshaanAgarwal[m]> <jonpsy[m]> "You could update us via your..." <- I have written most of the updates and I will sync that up with the latest progress.

12:59 <rcurtin[m]> @Guest786 if you trained the model with an mlpack binding you should be able to pickle or depickle the model to load it. it's easiest to do this if you create the model originally with the python bindings

13:02 <Guest781> If i create the model originally with c++ ,  how can i load to python binding for inferencing?

13:09 <Guest7887> Hi Team ,I am having mlpack model which is originally written in C++, how can i load the model to python for inferencing.

13:13 <Guest7860> Hi Team , I am having mlpack model which is originally written in C++, how can i load to python for inferencing?

13:13 <Guest7860> could you please help me on that

13:22 <zoq[m]> <EshaanAgarwal[m]> "Hi zoq ! Would it be possible to..." <- Hm, okay, I guess you are not available otherwise?

13:26 <EshaanAgarwal[m]> zoq[m]: Sorry, I can't understand what you mean here.

13:32 <EshaanAgarwal[m]> <EshaanAgarwal[m]> "Sorry, I can't understand what..." <- I am available but it would be difficult to gmeet on poor network. I would love to make some of the changes that you were talked abt earlier.

13:33 <EshaanAgarwal[m]> * I am available but it would be difficult to gmeet on poor network. I would love to make some of the changes that you were talked abt earlier and code.

14:51 <Guest7895> Hi Team i have a mlpack model which is originally written in C++, how can i load to python for inferencing ?

14:51 <Guest7895> could you please any one help me on that

15:03 <rcurtin[m]> Guest7895: I already answered the question but I don't think you saw the response:

15:03 <rcurtin[m]> > if you trained the model with an mlpack binding you should be able to pickle or depickle the model to load it. it's easiest to do this if you create the model originally with the python bindings

15:03 <rcurtin[m]> it could be tricky, if the model was written in C++; the Python interface to mlpack is all through the bindings, so if you have a way to retrain the model through the bindings, you will have a much easier time

15:12 <Guest7895> the question then would be if model is trained with python, can a c++  load the model ?

15:14 <rcurtin[m]> you would need to make sure that you know the exact C++ type produced by the binding, but yes, that should work

15:15 <rcurtin[m]> the `__getstate__()` function for a model should produce the exact serialized representation as a binary blob; you should be able to write that to a file, and then load it from C++ with `data::Load()`

15:15 <rcurtin[m]> there may be some tricky bits to get that right, but that should be the general idea; if you serialize with pickle, I think that will add some extra header information, so you need to get the exact serialized representation with `__getstate__()`

15:17 <rcurtin[m]> I think you can also use `_get_cpp_params()` to get the JSON representation of the model (which you can then write to file and load from C++), but JSON is bigger than the binary serialization, so that is a disadvantage of that approach

18:02 <Guest7895> I have a mlpack model which was written in C++, how can i retrain it using aws data bricks ?

18:02 <Guest7895> any idea suggestion pls

18:22 <zoq[m]> <Guest7895> "any idea suggestion pls" <- Have you seen the response above?

18:27 <Guest7895> no

18:29 <zoq[m]> https://libera.irclog.whitequark.org/mlpack/2022-09-28

18:29 <zoq[m]> Has the channel log