Cache transformed in QuantileDMatrix for efficiency. (#8666)

python-package/xgboost/core.py

@@ -407,21 +407,28 @@ class DataIter(ABC):  # pylint: disable=too-many-instance-attributes
 
     Parameters
     ----------
-    cache_prefix:
+    cache_prefix :
         Prefix to the cache files, only used in external memory.  It can be either an
         URI or a file path.
+    release_data :
+        Whether the iterator should release the data during reset. Set it to True if the
+        data transformation (converting data to np.float32 type) is expensive.
 
     """
 
-    def __init__(self, cache_prefix: Optional[str] = None) -> None:
+    def __init__(
+        self, cache_prefix: Optional[str] = None, release_data: bool = True
+    ) -> None:
         self.cache_prefix = cache_prefix
 
         self._handle = _ProxyDMatrix()
         self._exception: Optional[Exception] = None
         self._enable_categorical = False
         self._allow_host = True
+        self._release = release_data
         # Stage data in Python until reset or next is called to avoid data being free.
-        self._temporary_data: Optional[Tuple[Any, Any]] = None
+        self._temporary_data: Optional[Tuple[Any, Any, Any, Any]] = None
+        self._input_id: int = 0
 
     def get_callbacks(
         self, allow_host: bool, enable_categorical: bool
@@ -477,7 +484,8 @@ class DataIter(ABC):  # pylint: disable=too-many-instance-attributes
     def _reset_wrapper(self, this: None) -> None:  # pylint: disable=unused-argument
         """A wrapper for user defined `reset` function."""
         # free the data
-        self._temporary_data = None
+        if self._release:
+            self._temporary_data = None
         self._handle_exception(self.reset, None)
 
     def _next_wrapper(self, this: None) -> int:  # pylint: disable=unused-argument
@@ -498,20 +506,25 @@ class DataIter(ABC):  # pylint: disable=too-many-instance-attributes
         ) -> None:
             from .data import _proxy_transform, dispatch_proxy_set_data
 
-            new, cat_codes, feature_names, feature_types = _proxy_transform(
-                data,
-                feature_names,
-                feature_types,
-                self._enable_categorical,
-            )
+            # Reduce the amount of transformation that's needed for QuantileDMatrix.
+            if self._temporary_data is not None and id(data) == self._input_id:
+                new, cat_codes, feature_names, feature_types = self._temporary_data
+            else:
+                new, cat_codes, feature_names, feature_types = _proxy_transform(
+                    data,
+                    feature_names,
+                    feature_types,
+                    self._enable_categorical,
+                )
             # Stage the data, meta info are copied inside C++ MetaInfo.
-            self._temporary_data = (new, cat_codes)
+            self._temporary_data = (new, cat_codes, feature_names, feature_types)
             dispatch_proxy_set_data(self.proxy, new, cat_codes, self._allow_host)
             self.proxy.set_info(
                 feature_names=feature_names,
                 feature_types=feature_types,
                 **kwargs,
             )
+            self._input_id = id(data)
         # pylint: disable=not-callable
         return self._handle_exception(lambda: self.next(input_data), 0)
 

python-package/xgboost/data.py

@@ -1174,7 +1174,10 @@ class SingleBatchInternalIter(DataIter):  # pylint: disable=R0902
     def __init__(self, **kwargs: Any) -> None:
         self.kwargs = kwargs
         self.it = 0             # pylint: disable=invalid-name
-        super().__init__()
+
+        # This does not necessarily increase memory usage as the data transformation
+        # might use memory.
+        super().__init__(release_data=False)
 
     def next(self, input_data: Callable) -> int:
         if self.it == 1:

tests/python/test_data_iterator.py

@@ -1,4 +1,4 @@
-from typing import Dict, List
+from typing import Callable, Dict, List
 
 import numpy as np
 import pytest
@@ -153,3 +153,30 @@ def test_data_iterator(
     run_data_iterator(
         n_samples_per_batch, n_features, n_batches, "hist", subsample, False
     )
+
+
+class IterForCacheTest(xgb.DataIter):
+    def __init__(self, x: np.ndarray, y: np.ndarray, w: np.ndarray) -> None:
+        self.kwargs = {"data": x, "label": y, "weight": w}
+        super().__init__(release_data=False)
+
+    def next(self, input_data: Callable) -> int:
+        if self.it == 1:
+            return 0
+        self.it += 1
+        input_data(**self.kwargs)
+        return 1
+
+    def reset(self) -> None:
+        self.it = 0
+
+
+def test_data_cache() -> None:
+    n_batches = 1
+    n_features = 2
+    n_samples_per_batch = 16
+    data = make_batches(n_samples_per_batch, n_features, n_batches, False)
+    batches = [v[0] for v in data]
+    it = IterForCacheTest(*batches)
+    xgb.QuantileDMatrix(it)
+    assert it._input_id == id(batches[0])