Correctly handle missing categorical data in experimental FIL (#6132)

Correctly handle missing categorical data in FIL when 0 is one of the included categories.

Resolve #5578.

Authors:
  - William Hicks (https://github.com/wphicks)

Approvers:
  - Philip Hyunsu Cho (https://github.com/hcho3)
  - Dante Gama Dessavre (https://github.com/dantegd)

URL: #6132

cpp/include/cuml/experimental/fil/detail/evaluate_tree.hpp

@@ -65,7 +65,7 @@ HOST DEVICE auto evaluate_tree_impl(node_t const* __restrict__ node,
       if (cur_node.is_categorical()) {
         auto valid_categories = categorical_set_type{
           &cur_node.index(), uint32_t(sizeof(typename node_t::index_type) * 8)};
-        condition = valid_categories.test(input_val);
+        condition = valid_categories.test(input_val) && !isnan(input_val);
       } else {
         condition = (input_val < cur_node.threshold());
       }

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

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -870,3 +870,51 @@ def test_apply(train_device, infer_device, n_classes, tmp_path):
             expected_shape = (n_rows, num_boost_round * n_classes)
         assert pred_leaf.shape == expected_shape
         np.testing.assert_equal(pred_leaf, expected_pred_leaf)
+
+
+def test_missing_categorical():
+    builder = treelite.model_builder.ModelBuilder(
+        threshold_type="float32",
+        leaf_output_type="float32",
+        metadata=treelite.model_builder.Metadata(
+            num_feature=1,
+            task_type="kBinaryClf",
+            average_tree_output=False,
+            num_target=1,
+            num_class=[1],
+            leaf_vector_shape=(1, 1),
+        ),
+        tree_annotation=treelite.model_builder.TreeAnnotation(
+            num_tree=1, target_id=[0], class_id=[0]
+        ),
+        postprocessor=treelite.model_builder.PostProcessorFunc(
+            name="identity"
+        ),
+        base_scores=[0.0],
+    )
+    builder.start_tree()
+    builder.start_node(0)
+    builder.categorical_test(
+        feature_id=0,
+        category_list=[0, 2],
+        default_left=False,
+        category_list_right_child=False,
+        left_child_key=1,
+        right_child_key=2,
+    )
+    builder.end_node()
+    builder.start_node(1)
+    builder.leaf(1.0)
+    builder.end_node()
+    builder.start_node(2)
+    builder.leaf(2.0)
+    builder.end_node()
+    builder.end_tree()
+
+    model = builder.commit()
+
+    input = np.array([[np.nan]])
+    gtil_preds = treelite.gtil.predict(model, input)
+    fm = ForestInference.load_from_treelite_model(model)
+    fil_preds = np.asarray(fm.predict(input))
+    np.testing.assert_equal(fil_preds.flatten(), gtil_preds.flatten())