Add support for batched Iteration (#102)

* Add:Support for batched iteration

* Refactor:Data to be list of numpy arrays

* Add:test for tuple of numpy arrays

* Add:batched iteration support
* iter_batches function in Dataset class returns a BatchLoader object
* BatchLoader class added
* Moved utils.py
* Renamed utils.py
* Created test_data.py
* Cleanup
* Fix Typo

* Update src/sparsezoo/utils/data.py

Co-authored-by: Benjamin Fineran <[email protected]>

* Fix:Single-Input cases
Address:PR review comments

* Update:Rename tests/utils.py to tests/helpers.py
Fix:Unwrapping Single Input Errors

* Update:Rename tests/utils.py to tests/helpers.py
Fix:Unwrapping Single Input Errors
Update:tests_data.py

* Update:fixes from PR comments

Co-authored-by: Benjamin Fineran <[email protected]>

src/sparsezoo/utils/data.py

@@ -19,7 +19,7 @@
 import logging
 import math
 from collections import OrderedDict
-from typing import Dict, Iterable, Iterator, List, Tuple, Union
+from typing import Dict, Generator, Iterable, Iterator, List, Tuple, Union
 
 import numpy
 
@@ -28,10 +28,102 @@
 
 __all__ = ["Dataset", "RandomDataset", "DataLoader"]
 
-
 _LOGGER = logging.getLogger(__name__)
 
 
+#    A utility class to load data in batches for fixed number of iterations
+
+
+class _BatchLoader:
+    __slots__ = [
+        "_data",
+        "_batch_size",
+        "_was_wrapped_originally",
+        "_iterations",
+        "_batch_buffer",
+        "_batch_template",
+        "_batches_created",
+    ]
+
+    def __init__(
+        self,
+        data: Iterable[Union[numpy.ndarray, List[numpy.ndarray]]],
+        batch_size: int,
+        iterations: int,
+    ):
+        self._data = data
+        self._was_wrapped_originally = type(self._data[0]) is list
+        if not self._was_wrapped_originally:
+            self._data = [self._data]
+        self._batch_size = batch_size
+        self._iterations = iterations
+        if batch_size <= 0 or iterations <= 0:
+            raise ValueError(
+                f"Both batch size and number of iterations should be positive, "
+                f"supplied values (batch_size, iterations):{(batch_size, iterations)}"
+            )
+
+        self._batch_buffer = []
+        self._batch_template = self._init_batch_template()
+        self._batches_created = 0
+
+    def __iter__(self) -> Generator[List[numpy.ndarray], None, None]:
+        yield from self._multi_input_batch_generator()
+
+    @property
+    def _buffer_is_full(self) -> bool:
+        return len(self._batch_buffer) == self._batch_size
+
+    @property
+    def _all_batches_loaded(self) -> bool:
+        return self._batches_created >= self._iterations
+
+    def _multi_input_batch_generator(
+        self,
+    ) -> Generator[List[numpy.ndarray], None, None]:
+        # A generator for with each element of the form
+        # [[(batch_size, features_a), (batch_size, features_b), ...]]
+        while not self._all_batches_loaded:
+            yield from self._batch_generator(source=self._data)
+
+    def _batch_generator(self, source) -> Generator[List[numpy.ndarray], None, None]:
+        # batches from source
+        for sample in source:
+            self._batch_buffer.append(sample)
+            if self._buffer_is_full:
+                _batch = self._make_batch()
+                yield _batch
+                self._batch_buffer = []
+                self._batches_created += 1
+                if self._all_batches_loaded:
+                    break
+
+    def _init_batch_template(
+        self,
+    ) -> Iterable[Union[List[numpy.ndarray], numpy.ndarray]]:
+        # A placeholder for batches
+        return [
+            numpy.ascontiguousarray(
+                numpy.zeros((self._batch_size, *_input.shape), dtype=_input.dtype)
+            )
+            for _input in self._data[0]
+        ]
+
+    def _make_batch(self) -> Iterable[Union[numpy.ndarray, List[numpy.ndarray]]]:
+        # Copy contents of buffer to batch placeholder
+        # and return A list of numpy array(s) representing the batch
+
+        batch = [
+            numpy.stack([sample[idx] for sample in self._batch_buffer], out=template)
+            for idx, template in enumerate(self._batch_template)
+        ]
+
+        if not self._was_wrapped_originally:
+            # unwrap outer list
+            batch = batch[0]
+        return batch
+
+
 class Dataset(Iterable):
     """
     A numpy dataset implementation
@@ -76,6 +168,22 @@ def data(self) -> List[Union[numpy.ndarray, Dict[str, numpy.ndarray]]]:
         """
         return self._data
 
+    def iter_batches(
+        self, batch_size: int, iterations: int
+    ) -> Generator[List[numpy.ndarray], None, None]:
+        """
+        A function to iterate over data in batches
+
+        :param batch_size: non-negative integer representing the size of each
+        :param iterations: non-negative integer representing
+            the number of batches to return
+        :returns: A generator for batches, each batch is enclosed in a list
+            Each batch is of the form [(batch_size, *feature_shape)]
+        """
+        return _BatchLoader(
+            data=self.data, batch_size=batch_size, iterations=iterations
+        )
+
 
 class RandomDataset(Dataset):
     """

tests/sparsezoo/models/classification/test_efficientnet.py

@@ -15,7 +15,7 @@
 import pytest
 
 from sparsezoo.models.classification import efficientnet_b0, efficientnet_b4
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/classification/test_inception.py

@@ -15,7 +15,7 @@
 import pytest
 
 from sparsezoo.models.classification import inception_v3
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/classification/test_mobilenet.py

@@ -14,7 +14,7 @@
 import pytest
 
 from sparsezoo.models.classification import mobilenet_v1, mobilenet_v2
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/classification/test_resnet.py

@@ -23,7 +23,7 @@
     resnet_101_2x,
     resnet_152,
 )
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/classification/test_vgg.py

@@ -24,7 +24,7 @@
     vgg_19,
     vgg_19bn,
 )
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/detection/test_ssd.py

@@ -15,7 +15,7 @@
 import pytest
 
 from sparsezoo.models.detection import ssd_resnet50_300
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/detection/test_yolo.py

@@ -15,7 +15,7 @@
 import pytest
 
 from sparsezoo.models.detection import yolo_v3
-from tests.sparsezoo.utils import model_constructor
+from tests.sparsezoo.helpers import model_constructor
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/test_zoo.py

@@ -19,7 +19,7 @@
 
 from sparsezoo import Zoo
 from sparsezoo.utils import CACHE_DIR
-from tests.sparsezoo.utils import validate_downloaded_model
+from tests.sparsezoo.helpers import validate_downloaded_model
 
 
 @pytest.mark.parametrize(

tests/sparsezoo/models/test_zoo_extensive.py

@@ -17,7 +17,7 @@
 import pytest
 
 from sparsezoo.models import Zoo
-from tests.sparsezoo.utils import download_and_verify
+from tests.sparsezoo.helpers import download_and_verify
 
 
 def _get_models(domain, sub_domain) -> List[str]:

tests/sparsezoo/utils/__init__.py

@@ -0,0 +1,13 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.