From af2d3cdd20c56e14b1d859bfbde7443f0659d1ca Mon Sep 17 00:00:00 2001 From: Matteo Interlandi Date: Wed, 19 Aug 2020 09:12:43 -0700 Subject: [PATCH] add tweedie post_transform (#242) add tweedie to lgbm --- .../ml/operator_converters/constants.html | 10 +++++- .../onnx/onnx_operator.html | 36 ++++--------------- .../operator_converters/sklearn/lightgbm.html | 11 ++++-- .../operator_converters/sklearn/skl_sv.html | 12 ++----- .../ml/operator_converters/_gbdt_commons.py | 8 +++++ .../ml/operator_converters/constants.py | 3 ++ .../operator_converters/sklearn/lightgbm.py | 3 ++ tests/test_lightgbm_converter.py | 17 +++++++++ 8 files changed, 56 insertions(+), 44 deletions(-) diff --git a/doc/html/hummingbird/ml/operator_converters/constants.html b/doc/html/hummingbird/ml/operator_converters/constants.html index 22a324f1b..a1237f237 100644 --- a/doc/html/hummingbird/ml/operator_converters/constants.html +++ b/doc/html/hummingbird/ml/operator_converters/constants.html @@ -27,7 +27,7 @@

Module hummingbird.ml.operator_converters.constants Expand source code -Browse git +Browse git 
# -------------------------------------------------------------------------
 # Copyright (c) Microsoft Corporation. All rights reserved.
@@ -54,6 +54,9 @@ 
Module hummingbird.ml.operator_converters.constantsGlobal variables

 

 
The test input data for models that need to be traced

 

+
var TWEEDIE

+

+
Tweedie post transform

+

 
 
 

@@ -173,6 +180,7 @@ 
Index

 
  • SIGMOID
    • 
 
  • SOFTMAX
    • 
 
  • TEST_INPUT
    • 
+
  • TWEEDIE
    • 
 
 
 
diff --git a/doc/html/hummingbird/ml/operator_converters/onnx/onnx_operator.html b/doc/html/hummingbird/ml/operator_converters/onnx/onnx_operator.html
index a4d297863..5a387b069 100644
--- a/doc/html/hummingbird/ml/operator_converters/onnx/onnx_operator.html
+++ b/doc/html/hummingbird/ml/operator_converters/onnx/onnx_operator.html
@@ -327,18 +327,10 @@ 
    Ancestors
    
 
    Methods
    
 
    
 
    
-def forward(self, x)
+def forward(self, x) -> Callable[..., Any]
 
    
 
    
-
    Defines the computation performed at every call.
    
-
    Should be overridden by all subclasses.
    
-
    
-
    Note
    
-
    Although the recipe for forward pass needs to be defined within
-this function, one should call the :class:Module instance afterwards
-instead of this since the former takes care of running the
-registered hooks while the latter silently ignores them.
    
-
    
+
    
 
    
 
 Expand source code
@@ -377,18 +369,10 @@ 
    Ancestors
    
 
    Methods
    
 
    
 
    
-def forward(self, *x)
+def forward(self, *x) -> Callable[..., Any]
 
    
 
    
-
    Defines the computation performed at every call.
    
-
    Should be overridden by all subclasses.
    
-
    
-
    Note
    
-
    Although the recipe for forward pass needs to be defined within
-this function, one should call the :class:Module instance afterwards
-instead of this since the former takes care of running the
-registered hooks while the latter silently ignores them.
    
-
    
+
    
 
    
 
 Expand source code
@@ -429,18 +413,10 @@ 
    Ancestors
    
 
    Methods
    
 
    
 
    
-def forward(self, x)
+def forward(self, x) -> Callable[..., Any]
 
    
 
    
-
    Defines the computation performed at every call.
    
-
    Should be overridden by all subclasses.
    
-
    
-
    Note
    
-
    Although the recipe for forward pass needs to be defined within
-this function, one should call the :class:Module instance afterwards
-instead of this since the former takes care of running the
-registered hooks while the latter silently ignores them.
    
-
    
+
    
 
    
 
 Expand source code
diff --git a/doc/html/hummingbird/ml/operator_converters/sklearn/lightgbm.html b/doc/html/hummingbird/ml/operator_converters/sklearn/lightgbm.html
index e8cdb6fc8..bb98c478a 100644
--- a/doc/html/hummingbird/ml/operator_converters/sklearn/lightgbm.html
+++ b/doc/html/hummingbird/ml/operator_converters/sklearn/lightgbm.html
@@ -27,7 +27,7 @@ 
    Module hummingbird.ml.operator_converters.sklearn.lightg
 
    
 
 Expand source code
-Browse git
+Browse git
 
 
    # -------------------------------------------------------------------------
 # Copyright (c) Microsoft Corporation. All rights reserved.
@@ -42,6 +42,7 @@ 
    Module hummingbird.ml.operator_converters.sklearn.lightg
 import numpy as np
 from onnxconverter_common.registration import register_converter
 
+from .. import constants
 from .._gbdt_commons import convert_gbdt_classifier_common, convert_gbdt_common
 from .._tree_commons import TreeParameters
 
@@ -122,6 +123,8 @@ 
    Module hummingbird.ml.operator_converters.sklearn.lightg
     # Get tree information out of the model.
     n_features = operator.raw_operator._n_features
     tree_infos = operator.raw_operator.booster_.dump_model()["tree_info"]
+    if operator.raw_operator._objective == "tweedie":
+        extra_config[constants.POST_TRANSFORM] = constants.TWEEDIE
 
     return convert_gbdt_common(tree_infos, _get_tree_parameters, n_features, extra_config=extra_config)
 
@@ -161,7 +164,7 @@ 
    Returns
    
 
    
 
 Expand source code
-Browse git
+Browse git
 
 
    def convert_sklearn_lgbm_classifier(operator, device, extra_config):
     """
@@ -207,7 +210,7 @@ 
    Returns
    
 
    
 
 Expand source code
-Browse git
+Browse git
 
 
    def convert_sklearn_lgbm_regressor(operator, device, extra_config):
     """
@@ -226,6 +229,8 @@ 
    Returns
    
     # Get tree information out of the model.
     n_features = operator.raw_operator._n_features
     tree_infos = operator.raw_operator.booster_.dump_model()["tree_info"]
+    if operator.raw_operator._objective == "tweedie":
+        extra_config[constants.POST_TRANSFORM] = constants.TWEEDIE
 
     return convert_gbdt_common(tree_infos, _get_tree_parameters, n_features, extra_config=extra_config)
    
 
    
diff --git a/doc/html/hummingbird/ml/operator_converters/sklearn/skl_sv.html b/doc/html/hummingbird/ml/operator_converters/sklearn/skl_sv.html
index 647c79f7d..51e5ea058 100644
--- a/doc/html/hummingbird/ml/operator_converters/sklearn/skl_sv.html
+++ b/doc/html/hummingbird/ml/operator_converters/sklearn/skl_sv.html
@@ -304,18 +304,10 @@ 
    Ancestors
    
 
    Methods
    
 
    
 
    
-def forward(self, x)
+def forward(self, x) -> Callable[..., Any]
 
    
 
    
-
    Defines the computation performed at every call.
    
-
    Should be overridden by all subclasses.
    
-
    
-
    Note
    
-
    Although the recipe for forward pass needs to be defined within
-this function, one should call the :class:Module instance afterwards
-instead of this since the former takes care of running the
-registered hooks while the latter silently ignores them.
    
-
    
+
    
 
    
 
 Expand source code
diff --git a/hummingbird/ml/operator_converters/_gbdt_commons.py b/hummingbird/ml/operator_converters/_gbdt_commons.py
index 68cd5e8b2..7aab75742 100644
--- a/hummingbird/ml/operator_converters/_gbdt_commons.py
+++ b/hummingbird/ml/operator_converters/_gbdt_commons.py
@@ -124,6 +124,9 @@ def apply_sigmoid(x):
     def apply_softmax(x):
         return torch.softmax(x, dim=1)
 
+    def apply_tweedie(x):
+        return torch.exp(x)
+
     # For models following the Sklearn API we need to build the post transform ourselves.
     if classes is not None and constants.POST_TRANSFORM not in extra_config:
         if len(classes) <= 2:
@@ -143,6 +146,11 @@ def apply_softmax(x):
                 extra_config[constants.POST_TRANSFORM] = lambda x: apply_softmax(apply_base_prediction(base_prediction)(x))
             else:
                 extra_config[constants.POST_TRANSFORM] = apply_softmax
+        elif extra_config[constants.POST_TRANSFORM] == constants.TWEEDIE:
+            if constants.BASE_PREDICTION in extra_config:
+                extra_config[constants.POST_TRANSFORM] = lambda x: apply_tweedie(apply_base_prediction(base_prediction)(x))
+            else:
+                extra_config[constants.POST_TRANSFORM] = apply_tweedie
         else:
             raise NotImplementedError("Post transform {} not implemeneted yet".format(extra_config[constants.POST_TRANSFORM]))
     elif constants.BASE_PREDICTION in extra_config:
diff --git a/hummingbird/ml/operator_converters/constants.py b/hummingbird/ml/operator_converters/constants.py
index a03bf9633..4c1f7219a 100644
--- a/hummingbird/ml/operator_converters/constants.py
+++ b/hummingbird/ml/operator_converters/constants.py
@@ -23,6 +23,9 @@
 SOFTMAX = "SOFTMAX"
 """Softmax post transform"""
 
+TWEEDIE = "TWEEDIE"
+"""Tweedie post transform"""
+
 GET_PARAMETERS_FOR_TREE_TRAVERSAL = "get_parameters_for_tree_trav"
 """Which function to use to extract the parameters for the tree traversal strategy"""
 
diff --git a/hummingbird/ml/operator_converters/sklearn/lightgbm.py b/hummingbird/ml/operator_converters/sklearn/lightgbm.py
index f6de6b258..b51847e15 100644
--- a/hummingbird/ml/operator_converters/sklearn/lightgbm.py
+++ b/hummingbird/ml/operator_converters/sklearn/lightgbm.py
@@ -11,6 +11,7 @@
 import numpy as np
 from onnxconverter_common.registration import register_converter
 
+from .. import constants
 from .._gbdt_commons import convert_gbdt_classifier_common, convert_gbdt_common
 from .._tree_commons import TreeParameters
 
@@ -91,6 +92,8 @@ def convert_sklearn_lgbm_regressor(operator, device, extra_config):
     # Get tree information out of the model.
     n_features = operator.raw_operator._n_features
     tree_infos = operator.raw_operator.booster_.dump_model()["tree_info"]
+    if operator.raw_operator._objective == "tweedie":
+        extra_config[constants.POST_TRANSFORM] = constants.TWEEDIE
 
     return convert_gbdt_common(tree_infos, _get_tree_parameters, n_features, extra_config=extra_config)
 
diff --git a/tests/test_lightgbm_converter.py b/tests/test_lightgbm_converter.py
index c51a004da..16954887b 100644
--- a/tests/test_lightgbm_converter.py
+++ b/tests/test_lightgbm_converter.py
@@ -243,6 +243,23 @@ def test_lgbm_classifier_random_forest(self):
         self.assertIsNotNone(torch_model)
         np.testing.assert_allclose(model.predict_proba(X), torch_model.predict_proba(X), rtol=1e-06, atol=1e-06)
 
+    # Test Tweedie loss in lgbm
+    @unittest.skipIf(not lightgbm_installed(), reason="LightGBM test requires LightGBM installed")
+    def test_lgbm_tweedie(self):
+        warnings.filterwarnings("ignore")
+        model = lgb.LGBMRegressor(objective="tweedie", n_estimators=2, max_depth=5)
+
+        np.random.seed(0)
+        X = np.random.rand(100, 200)
+        X = np.array(X, dtype=np.float32)
+        y = np.random.randint(100, size=100)
+
+        model.fit(X, y)
+
+        torch_model = hummingbird.ml.convert(model, "torch")
+        self.assertIsNotNone(torch_model)
+        np.testing.assert_allclose(model.predict(X), torch_model.predict(X), rtol=1e-06, atol=1e-06)
+
     # Backend tests.
     # Test TorchScript backend regression.
     @unittest.skipIf(not lightgbm_installed(), reason="LightGBM test requires LightGBM installed")