-
Notifications
You must be signed in to change notification settings - Fork 3.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CoreML support for LightGBM models #1074
Comments
@karthikv2k Probably not, GBDT has mostly nothing to do with tensor programming, logical (boolean) tensors for computational graphs are not appropriate (both in speed and efficiency) versus the current implementation of prediction. Unless someone generates a converter from tree ensembles to neural network / computational graph. |
@Laurae2 Thanks for your prompt response. I am looking for a standard way to export LightGBM models into a language agnostic format. This will allow us to build production ML inferencing systems that can take this format as input. For example, CoreML provides a way to convert XGBoost models to their open CoreML protobuf based spec. I see there is a link to JPMML in the readme to convert LightGBM into PMML format but JPMML has the viral AGPL licensing. |
I just realized there is a PMML convertor in pmml/pmml.py . It is good to have CoreML support too similar to XGBoost and CatBoost . I think I can work on it. |
@karthikv2k Thanks. here is another solution: https://github.com/dmlc/treelite |
@wxchan @StrikerRUS what is your thoughts about CoreML support ? |
@guolinke I've never hear about CoreML before this issue, because I'm far away from Apple world.
However, if @karthikv2k have time and interest in developing it, then it's good and valuable enhancement for LightGBM. |
@karthikv2k Any news? Guys from ONNX have added support for LightGBM (only |
AFAIK ML.NET uses LightGBM internally and supports ONNX export. |
I used xgboost to coreml conversion, the problem is that the inputs for converted mlmodel are so that we should input features separately! (not sure if we can enter features as a single vector). |
Closed in favor of being in #2302. We decided to keep all feature requests in one place. Welcome to contribute this feature! Please re-open this issue (or post a comment if you are not a topic starter) if you are actively working on implementing this feature. |
Open PR for adding LightGBM to CoreML: apple/coremltools#254. It was opened on Oct 4 2018 but still is not merged. Someone with the CoreML knowledge may help to review and push that PR forward. XGBoost converter: https://github.com/apple/coremltools/tree/master/coremltools/converters/xgboost. |
Is there a plan to support ONNX as an output format for LightGBM models?
The text was updated successfully, but these errors were encountered: