Update FIL tests to use XGBoost UBJSON instead of binary (#6153)

Starting from 2.1, XGBoost uses UBJSON format to serialize models.
Replace all uses of the legacy model format with UBJSON.

Also make `xgboost` a test dependency of cuML so that the FIL tests run in the CI pipelines.

Authors:
  - Philip Hyunsu Cho (https://github.com/hcho3)

Approvers:
  - William Hicks (https://github.com/wphicks)
  - Robert Maynard (https://github.com/robertmaynard)
  - https://github.com/jakirkham

URL: #6153

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -77,4 +77,5 @@ dependencies:
 - sysroot_linux-64==2.17
 - treelite==4.3.0
 - umap-learn==0.5.6
+- xgboost>=2.1.0
 name: all_cuda-118_arch-x86_64

conda/environments/all_cuda-125_arch-x86_64.yaml

@@ -73,4 +73,5 @@ dependencies:
 - sysroot_linux-64==2.17
 - treelite==4.3.0
 - umap-learn==0.5.6
+- xgboost>=2.1.0
 name: all_cuda-125_arch-x86_64

dependencies.yaml

@@ -523,6 +523,7 @@ dependencies:
           - pytest-xdist
           - seaborn
           - *scikit_learn
+          - &xgboost xgboost>=2.1.0
           - statsmodels
           - umap-learn==0.5.6
           - pynndescent
@@ -537,4 +538,4 @@ dependencies:
           - pandas
           - *scikit_learn
           - seaborn
-          - xgboost
+          - *xgboost

python/cuml/cuml/tests/experimental/test_filex.py

@@ -163,7 +163,7 @@ def test_fil_classification(
             X, y, train_size=0.8, random_state=0
         )
 
-        model_path = os.path.join(tmp_path, "xgb_class.model")
+        model_path = os.path.join(tmp_path, "xgb_class.ubj")
 
         bst = _build_and_save_xgboost(
             model_path,
@@ -258,7 +258,7 @@ def test_fil_regression(
             X, y, train_size=train_size, random_state=0
         )
 
-        model_path = os.path.join(tmp_path, "xgb_reg.model")
+        model_path = os.path.join(tmp_path, "xgb_reg.ubj")
         bst = _build_and_save_xgboost(
             model_path,
             X_train,
@@ -490,12 +490,12 @@ def test_fil_skl_regression(
         np.testing.assert_allclose(fil_preds_opt, fil_preds, atol=1.2e-3)
 
 
-@pytest.fixture(scope="session", params=["binary", "json"])
+@pytest.fixture(scope="session", params=["ubjson", "json"])
 def small_classifier_and_preds(tmpdir_factory, request):
     X, y = simulate_data(500, 10, random_state=43210, classification=True)
 
-    ext = "json" if request.param == "json" else "model"
-    model_type = "xgboost_json" if request.param == "json" else "xgboost"
+    ext = "json" if request.param == "json" else "ubj"
+    model_type = "xgboost_json" if request.param == "json" else "xgboost_ubj"
     model_path = str(
         tmpdir_factory.mktemp("models").join(f"small_class.{ext}")
     )
@@ -738,7 +738,7 @@ def test_predict_per_tree(
             classification=True,
         )
 
-        model_path = os.path.join(tmp_path, "xgb_class.model")
+        model_path = os.path.join(tmp_path, "xgb_class.ubj")
 
         xgboost_params = {"base_score": (0.5 if n_classes == 2 else 0.0)}
         bst = _build_and_save_xgboost(
@@ -751,7 +751,7 @@ def test_predict_per_tree(
             xgboost_params=xgboost_params,
         )
         fm = ForestInference.load(model_path, output_class=True)
-        tl_model = treelite.Model.from_xgboost(bst)
+        tl_model = treelite.frontend.from_xgboost(bst)
         pred_per_tree_tl = treelite.gtil.predict_per_tree(tl_model, X)
 
     with using_device_type(infer_device):
@@ -773,7 +773,7 @@ def test_predict_per_tree(
         assert pred_per_tree.shape == expected_shape
         np.testing.assert_almost_equal(sum_by_class, margin_pred, decimal=3)
         np.testing.assert_almost_equal(
-            pred_per_tree, pred_per_tree_tl, decimal=3
+            pred_per_tree.reshape((n_rows, -1, 1)), pred_per_tree_tl, decimal=3
         )
         np.testing.assert_almost_equal(
             pred_per_tree_opt, pred_per_tree, decimal=3
@@ -844,7 +844,7 @@ def test_apply(train_device, infer_device, n_classes, tmp_path):
             classification=True,
         )
 
-        model_path = os.path.join(tmp_path, "xgb_class.model")
+        model_path = os.path.join(tmp_path, "xgb_class.ubj")
 
         xgboost_params = {"base_score": (0.5 if n_classes == 2 else 0.0)}
         bst = _build_and_save_xgboost(
@@ -858,7 +858,7 @@ def test_apply(train_device, infer_device, n_classes, tmp_path):
         )
 
         fm = ForestInference.load(
-            model_path, output_class=True, model_type="xgboost"
+            model_path, output_class=True, model_type="xgboost_ubj"
         )
 
     with using_device_type(infer_device):

python/cuml/cuml/tests/test_fil.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2019-2023, NVIDIA CORPORATION.
+# Copyright (c) 2019-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -155,7 +155,7 @@ def test_fil_classification(
         X, y, train_size=train_size, random_state=0
     )
 
-    model_path = os.path.join(tmp_path, "xgb_class.model")
+    model_path = os.path.join(tmp_path, "xgb_class.ubj")
 
     bst = _build_and_save_xgboost(
         model_path,
@@ -226,7 +226,7 @@ def test_fil_regression(n_rows, n_columns, num_rounds, tmp_path, max_depth):
         X, y, train_size=train_size, random_state=0
     )
 
-    model_path = os.path.join(tmp_path, "xgb_reg.model")
+    model_path = os.path.join(tmp_path, "xgb_reg.ubj")
     bst = _build_and_save_xgboost(
         model_path,
         X_train,
@@ -447,12 +447,12 @@ def test_fil_skl_regression(
     assert np.allclose(fil_preds, skl_preds, 1.2e-2)
 
 
-@pytest.fixture(scope="session", params=["binary", "json"])
+@pytest.fixture(scope="session", params=["ubjson", "json"])
 def small_classifier_and_preds(tmpdir_factory, request):
     X, y = simulate_data(500, 10, random_state=43210, classification=True)
 
-    ext = "json" if request.param == "json" else "model"
-    model_type = "xgboost_json" if request.param == "json" else "xgboost"
+    ext = "json" if request.param == "json" else "ubj"
+    model_type = "xgboost_json" if request.param == "json" else "xgboost_ubj"
     model_path = str(
         tmpdir_factory.mktemp("models").join(f"small_class.{ext}")
     )

python/cuml/pyproject.toml

@@ -137,6 +137,7 @@ test = [
     "seaborn",
     "statsmodels",
     "umap-learn==0.5.6",
+    "xgboost>=2.1.0",
 ] # This list was generated by `rapids-dependency-file-generator`. To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
 
 [project.urls]