Add convergence test for sparse datasets. (#7922)

tests/python-gpu/test_gpu_updaters.py

@@ -53,6 +53,23 @@ class TestGPUUpdaters:
         note(result)
         assert tm.non_increasing(result["train"][dataset.metric])
 
+    @given(tm.sparse_datasets_strategy)
+    @settings(deadline=None, print_blob=True)
+    def test_sparse(self, dataset):
+        param = {"tree_method": "hist", "max_bin": 64}
+        hist_result = train_result(param, dataset.get_dmat(), 16)
+        note(hist_result)
+        assert tm.non_increasing(hist_result['train'][dataset.metric])
+
+        param = {"tree_method": "gpu_hist", "max_bin": 64}
+        gpu_hist_result = train_result(param, dataset.get_dmat(), 16)
+        note(gpu_hist_result)
+        assert tm.non_increasing(gpu_hist_result['train'][dataset.metric])
+
+        np.testing.assert_allclose(
+            hist_result["train"]["rmse"], gpu_hist_result["train"]["rmse"], rtol=1e-2
+        )
+
     @given(strategies.integers(10, 400), strategies.integers(3, 8),
            strategies.integers(1, 2), strategies.integers(4, 7))
     @settings(deadline=None, print_blob=True)

tests/python/test_updaters.py

@@ -99,6 +99,23 @@ class TestTreeMethod:
         note(result)
         assert tm.non_increasing(result['train'][dataset.metric])
 
+    @given(tm.sparse_datasets_strategy)
+    @settings(deadline=None, print_blob=True)
+    def test_sparse(self, dataset):
+        param = {"tree_method": "hist", "max_bin": 64}
+        hist_result = train_result(param, dataset.get_dmat(), 16)
+        note(hist_result)
+        assert tm.non_increasing(hist_result['train'][dataset.metric])
+
+        param = {"tree_method": "approx", "max_bin": 64}
+        approx_result = train_result(param, dataset.get_dmat(), 16)
+        note(approx_result)
+        assert tm.non_increasing(approx_result['train'][dataset.metric])
+
+        np.testing.assert_allclose(
+            hist_result["train"]["rmse"], approx_result["train"]["rmse"]
+        )
+
     def test_hist_categorical(self):
         # hist must be same as exact on all-categorial data
         dpath = 'demo/data/'

tests/python/testing.py

@@ -1,5 +1,7 @@
-# coding: utf-8
+from concurrent.futures import ThreadPoolExecutor
 import os
+import multiprocessing
+from typing import Tuple, Union
 import urllib
 import zipfile
 import sys
@@ -11,6 +13,7 @@ import pytest
 import gc
 import xgboost as xgb
 import numpy as np
+from scipy import sparse
 import platform
 
 hypothesis = pytest.importorskip('hypothesis')
@@ -327,6 +330,127 @@ def make_categorical(
     return df, label
 
 
+@memory.cache
+def make_sparse_regression(
+    n_samples: int, n_features: int, sparsity: float, as_dense: bool
+) -> Tuple[Union[sparse.csr_matrix], np.ndarray]:
+    """Make sparse matrix.
+
+    Parameters
+    ----------
+
+    as_dense:
+
+      Return the matrix as np.ndarray with missing values filled by NaN
+
+    """
+    if not hasattr(np.random, "default_rng"):
+        # old version of numpy on s390x
+        rng = np.random.RandomState(1994)
+        X = sparse.random(
+            m=n_samples,
+            n=n_features,
+            density=1.0 - sparsity,
+            random_state=rng,
+            format="csr",
+        )
+        y = rng.normal(loc=0.0, scale=1.0, size=n_samples)
+        return X, y
+
+    # Use multi-thread to speed up the generation, convenient if you use this function
+    # for benchmarking.
+    n_threads = multiprocessing.cpu_count()
+    n_threads = min(n_threads, n_features)
+
+    def random_csc(t_id: int) -> sparse.csc_matrix:
+        rng = np.random.default_rng(1994 * t_id)
+        thread_size = n_features // n_threads
+        if t_id == n_threads - 1:
+            n_features_tloc = n_features - t_id * thread_size
+        else:
+            n_features_tloc = thread_size
+
+        X = sparse.random(
+            m=n_samples,
+            n=n_features_tloc,
+            density=1.0 - sparsity,
+            random_state=rng,
+        ).tocsc()
+        y = np.zeros((n_samples, 1))
+
+        for i in range(X.shape[1]):
+            size = X.indptr[i + 1] - X.indptr[i]
+            if size != 0:
+                y += X[:, i].toarray() * rng.random((n_samples, 1)) * 0.2
+
+        return X, y
+
+    futures = []
+    with ThreadPoolExecutor(max_workers=n_threads) as executor:
+        for i in range(n_threads):
+            futures.append(executor.submit(random_csc, i))
+
+    X_results = []
+    y_results = []
+    for f in futures:
+        X, y = f.result()
+        X_results.append(X)
+        y_results.append(y)
+
+    assert len(y_results) == n_threads
+
+    csr: sparse.csr_matrix = sparse.hstack(X_results, format="csr")
+    y = np.asarray(y_results)
+    y = y.reshape((y.shape[0], y.shape[1])).T
+    y = np.sum(y, axis=1)
+
+    assert csr.shape[0] == n_samples
+    assert csr.shape[1] == n_features
+    assert y.shape[0] == n_samples
+
+    if as_dense:
+        arr = csr.toarray()
+        arr[arr == 0] = np.nan
+        return arr, y
+
+    return csr, y
+
+
+sparse_datasets_strategy = strategies.sampled_from(
+    [
+        TestDataset(
+            "1e5x8-0.95-csr",
+            lambda: make_sparse_regression(int(1e5), 8, 0.95, False),
+            "reg:squarederror",
+            "rmse",
+        ),
+        TestDataset(
+            "1e5x8-0.5-csr",
+            lambda: make_sparse_regression(int(1e5), 8, 0.5, False),
+            "reg:squarederror",
+            "rmse",
+        ),
+        TestDataset(
+            "1e5x8-0.5-dense",
+            lambda: make_sparse_regression(int(1e5), 8, 0.5, True),
+            "reg:squarederror",
+            "rmse",
+        ),
+        TestDataset(
+            "1e5x8-0.05-csr",
+            lambda: make_sparse_regression(int(1e5), 8, 0.05, False),
+            "reg:squarederror",
+            "rmse",
+        ),
+        TestDataset(
+            "1e5x8-0.05-dense",
+            lambda: make_sparse_regression(int(1e5), 8, 0.05, True),
+            "reg:squarederror",
+            "rmse",
+        ),
+    ]
+)
+
 _unweighted_datasets_strategy = strategies.sampled_from(
     [
         TestDataset(