How to load python-trained LGB model file into LightGbmBinaryTrainer object, and use the trained model inside c#?

galeese commented 4 years ago

System information

OS version/distro: Windows 10
.NET Version (eg., dotnet --info):

Issue

What did you do? I've trained a LGB-model from the python lightgbm package.
What happened? I've obtained: model-file - "model.txt" file of model-parameters: "predict.conf"
What did you expect? I expect to somehow to create a LightGbmBinaryTrainer object based on the two files: "model.txt" and "predict.conf"

I don't want to run "./lightgbm" config=predict.conf from c# each time in order to obtain the new prediciton inside the *.txt file. I want to apply the created model inside the c# environment and get the prediction inside the c# and not to load it from the txt file.

Source code / logs

The contents of the "model.txt": tree version=v3 num_class=1 num_tree_per_iteration=1 label_index=0 max_feature_idx=10307 objective=binary sigmoid:1 feature_names=Column_0 Column_1 Column_2 Column_3 Column_4

The contents of the "predict.conf" task = predict

data = sharp_preds_string.csv

input_model =model.txt

output_result= LightGBM_predict_result.txt

Please paste or attach the code or logs or traces that would be helpful to diagnose the issue you are reporting.

galeese commented 4 years ago

transfer.zip

michaelgsharp commented 4 years ago

ML.Net can only import models that are either ML.Net models or ONNX models, so you can't load the model generated by the python LBG package. I looked at the documentation for python LBG and didn't see a way to export an onnx model either, though its possible I missed it. Is there a reason you don't want to directly create the model with ML.Net?

galeese commented 4 years ago

" Is there a reason you don't want to directly create the model with ML.Net?" Thank you Michael, what you intend - is the only thing left to do!)))

michaelgsharp commented 4 years ago

Sounds good. Since you have that answer I will be closing this issue for now. If you have any more questions about it please feel free to re-open this issue. Thanks!

galeese commented 4 years ago

agree to close

вт, 25 авг. 2020 г., 19:49 Michael Sharp notifications@github.com:

Sounds good. Since you have that answer I will be closing this issue for now. If you have any more questions about it please feel free to re-open this issue. Thanks!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/dotnet/machinelearning/issues/5360#issuecomment-680144035, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACUU2O24SK45EWHJ5DNHVZTSCPTR7ANCNFSM4QIDUN4Q .

dotnet / machinelearning