Fix Hummingbird API changes (#418)

* Hummingbird integration with lgbm
Add additional tests for Hummingbird

* Address flake8 errors

* we get hummingbird directly from github

* manually install torch

* roll back previous case

* skip HB tests for older versions of ORT

* fix reason and turn onnx into onnxruntime

* Addressing wenbingl comments

* fix to the git link

* add -f option to get hb

* add finx_links for pytorch in requirements
remove hummingbird_installed from utils
point to the onnxconveter-common branch with the hummingb_installed code

* Update convert.py

trigger pipeline

* point to the actual onnxconverter-common repository

* add torch to dev requirements
return the model from the hummingbird container

onnxmltools/convert/lightgbm/convert.py

@@ -65,7 +65,7 @@ def convert(model, name=None, initial_types=None, doc_string='', target_opset=No
         extra_config[constants.ONNX_INITIAL_TYPES] = initial_types
         extra_config[constants.ONNX_OUTPUT_MODEL_NAME] = name
         extra_config[constants.ONNX_TARGET_OPSET] = target_opset
-        onnx_model = convert(onnx_ml_model, "onnx", extra_config=extra_config)
+        onnx_model = convert(onnx_ml_model, "onnx", extra_config=extra_config).model
 
         return onnx_model
 

requirements-dev.txt

@@ -18,4 +18,5 @@ wheel
 xgboost
 catboost
 flake8
+torch==1.5.1+cpu
 hummingbird-ml