Merge pull request #2538 from hcho3/remove_protobuf_dep

[DISCUSS] Remove Protobuf dependency

CHANGELOG.md

@@ -48,6 +48,7 @@
 - PR #2513: Fixing LGTM Analysis Issues
 - PR #2099: Raise an error when float64 data is used with dask RF
 - PR #2526: Removing PCA TSQR as a solver due to scalability issues
+- PR #2538: Remove Protobuf dependency
 
 ## Bug Fixes
 - PR #2369: Update RF code to fix set_params memory leak

ci/mg/build.sh

@@ -51,7 +51,6 @@ conda install -c conda-forge -c rapidsai -c rapidsai-nightly -c nvidia \
       "lapack" \
       "cmake==3.14.3" \
       "umap-learn" \
-      "protobuf>=3.4.1,<4.0.0" \
       "nccl>=2.5" \
       "dask>=2.12.0" \
       "distributed>=2.12.0" \

conda/environments/cuml_dev_cuda10.0.yml

@@ -28,7 +28,6 @@ dependencies:
 - nccl>=2.5
 - libcumlprims=0.15.0a200622
 - statsmodels
-- protobuf >=3.4.1,<4.0.0
 - treelite=0.92
 - doxygen
 - sphinx

conda/environments/cuml_dev_cuda10.1.yml

@@ -28,7 +28,6 @@ dependencies:
 - nccl>=2.5
 - libcumlprims=0.15.0a200622
 - statsmodels
-- protobuf >=3.4.1,<4.0.0
 - treelite=0.92
 - doxygen
 - sphinx

conda/environments/cuml_dev_cuda10.2.yml

@@ -28,7 +28,6 @@ dependencies:
 - nccl>=2.5
 - libcumlprims=0.15.0a200622
 - statsmodels
-- protobuf >=3.4.1,<4.0.0
 - treelite=0.92
 - doxygen
 - sphinx

conda/recipes/cuml/meta.yaml

@@ -28,7 +28,6 @@ requirements:
     - setuptools
     - cython>=0.29,<0.30
     - cmake>=3.14
-    - protobuf >=3.4.1,<4.0.0
     - treelite=0.92
     - cudf {{ minor_version }}
     - libcuml={{ version }}

conda/recipes/libcuml/meta.yaml

@@ -36,7 +36,6 @@ requirements:
     - ucx-py {{ minor_version }}
     - libcumlprims=0.15.0a200622
     - lapack
-    - protobuf >=3.4.1,<4.0.0
     - treelite=0.92
   run:
     - libcumlprims=0.15.0a200622

cpp/CMakeLists.txt

@@ -181,9 +181,6 @@ else()
   message(STATUS "Manually setting BLAS to ${BLAS_LIBRARIES}")
 endif()
 
-set(Protobuf_USE_STATIC_LIBS ON)
-find_package(Protobuf REQUIRED)
-
 set(CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
 
 ##############################################################################
@@ -316,7 +313,6 @@ set(CUML_PUBLIC_LINK_LIBRARIES
   )
 
 set(CUML_PRIVATE_LINK_LIBRARIES
-  ${Protobuf_LIBRARIES}
   faisslib
   treelite::treelite
   treelite::treelite_runtime

notebooks/forest_inference_demo.ipynb

@@ -5,9 +5,9 @@
    "metadata": {},
    "source": [
     "# Forest Inference Library (FIL)\n",
-    "The forest inference library is used to load saved forest models of xgboost, lightgbm or protobuf and perform inference on them. It can be used to perform both classification and regression. In this notebook, we'll begin by fitting a model with XGBoost and saving it. We'll then load the saved model into FIL and use it to infer on new data.\n",
+    "The forest inference library is used to load saved forest models of xgboost, lightgbm and perform inference on them. It can be used to perform both classification and regression. In this notebook, we'll begin by fitting a model with XGBoost and saving it. We'll then load the saved model into FIL and use it to infer on new data.\n",
     "\n",
-    "FIL works in the same way with lightgbm and protobuf model as well.\n",
+    "FIL works in the same way with lightgbm model as well.\n",
     "\n",
     "The model accepts both numpy arrays and cuDF dataframes. In order to convert your dataset to cudf format please read the cudf documentation on https://docs.rapids.ai/api/cudf/stable. \n",
     "\n",
@@ -225,7 +225,7 @@
     "                                    multiple rows per thread block\n",
     "        model_type : str\n",
     "            Format of saved treelite model to load.\n",
-    "            Can be 'xgboost', 'lightgbm', or 'protobuf'"
+    "            Can be 'xgboost', 'lightgbm'"
    ]
   },
   {

python/README.md

@@ -57,7 +57,6 @@ To build cuML's Python package, the following dependencies are required:
 - cython >= 0.29, < 0.30
 - numpy
 - cmake >= 3.14
-- protobuf >= 3.4.1, < 4.0.0
 - cudf version matching the cuML version
 - libcuml version matching the cuML version
 - libcuml={{ version }}

python/cuml/ensemble/randomforestclassifier.pyx

@@ -434,7 +434,7 @@ class RandomForestClassifier(Base, ClassifierMixin):
     def _get_serialized_model(self):
         """
         Returns the self.treelite_serialized_model.
-        Cuml RF model gets converted to treelite protobuf bytes by:
+        Cuml RF model gets converted to treelite bytes by:
             1. converting the cuml RF model to a treelite model. The treelite
             models handle (pointer) is returned
             2. The treelite model handle is converted to bytes.

python/cuml/ensemble/randomforestregressor.pyx

@@ -414,7 +414,7 @@ class RandomForestRegressor(Base, RegressorMixin):
     def _get_serialized_model(self):
         """
         Returns the self.treelite_serialized_model.
-        Cuml RF model gets converted to treelite protobuf bytes by:
+        Cuml RF model gets converted to treelite bytes by:
             1. converting the cuml RF model to a treelite model. The treelite
             models handle (pointer) is returned
             2. The treelite model handle is converted to bytes.

python/cuml/fil/fil.pyx

@@ -125,7 +125,7 @@ cdef class TreeliteModel():
             Path to treelite model file to load
 
         model_type : string
-            Type of model: 'xgboost', 'protobuf', or 'lightgbm'
+            Type of model: 'xgboost', or 'lightgbm'
         """
         filename_bytes = filename.encode("UTF-8")
         cdef ModelHandle handle
@@ -134,12 +134,6 @@ cdef class TreeliteModel():
             if res < 0:
                 err = TreeliteGetLastError()
                 raise RuntimeError("Failed to load %s (%s)" % (filename, err))
-        elif model_type == "protobuf":
-            # XXX Not tested
-            res = TreeliteLoadProtobufModel(filename_bytes, &handle)
-            if res < 0:
-                err = TreeliteGetLastError()
-                raise RuntimeError("Failed to load %s (%s)" % (filename, err))
         elif model_type == "lightgbm":
             res = TreeliteLoadLightGBMModel(filename_bytes, &handle)
             if res < 0:
@@ -620,7 +614,7 @@ class ForestInference(Base):
             See documentation in FIL.load_from_treelite_model
         model_type : string (default="xgboost")
             Format of the saved treelite model to be load.
-            It can be 'xgboost', 'lightgbm', or 'protobuf'.
+            It can be 'xgboost', 'lightgbm'.
 
         Returns
         ----------