!20047 [assistant][ops]New operator implementation, include QmnistDataset

Merge pull request !20047 from Wangsong95/qmnist_dataset

mindspore/ccsrc/minddata/dataset/api/datasets.cc

@@ -102,6 +102,7 @@
 #include "minddata/dataset/engine/ir/datasetops/source/text_file_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/manifest_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/minddata_node.h"
+#include "minddata/dataset/engine/ir/datasetops/source/qmnist_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/sbu_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/tf_record_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/usps_node.h"
@@ -1220,6 +1221,28 @@ MnistDataset::MnistDataset(const std::vector<char> &dataset_dir, const std::vect
 }
 
 #ifndef ENABLE_ANDROID
+QMnistDataset::QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                             const std::shared_ptr<Sampler> &sampler, const std::shared_ptr<DatasetCache> &cache) {
+  auto sampler_obj = sampler ? sampler->Parse() : nullptr;
+  auto ds = std::make_shared<QMnistNode>(CharToString(dataset_dir), CharToString(usage), compat, sampler_obj, cache);
+  ir_node_ = std::static_pointer_cast<DatasetNode>(ds);
+}
+
+QMnistDataset::QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                             const Sampler *sampler, const std::shared_ptr<DatasetCache> &cache) {
+  auto sampler_obj = sampler ? sampler->Parse() : nullptr;
+  auto ds = std::make_shared<QMnistNode>(CharToString(dataset_dir), CharToString(usage), compat, sampler_obj, cache);
+  ir_node_ = std::static_pointer_cast<DatasetNode>(ds);
+}
+
+QMnistDataset::QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                             const std::reference_wrapper<Sampler> sampler,
+                             const std::shared_ptr<DatasetCache> &cache) {
+  auto sampler_obj = sampler.get().Parse();
+  auto ds = std::make_shared<QMnistNode>(CharToString(dataset_dir), CharToString(usage), compat, sampler_obj, cache);
+  ir_node_ = std::static_pointer_cast<DatasetNode>(ds);
+}
+
 TextFileDataset::TextFileDataset(const std::vector<std::vector<char>> &dataset_files, int64_t num_samples,
                                  ShuffleMode shuffle, int32_t num_shards, int32_t shard_id,
                                  const std::shared_ptr<DatasetCache> &cache) {

mindspore/ccsrc/minddata/dataset/api/python/bindings/dataset/engine/ir/datasetops/source/bindings.cc

@@ -44,6 +44,7 @@
 #ifndef ENABLE_ANDROID
 #include "minddata/dataset/engine/ir/datasetops/source/manifest_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/minddata_node.h"
+#include "minddata/dataset/engine/ir/datasetops/source/qmnist_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/sbu_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/tf_record_node.h"
 #include "minddata/dataset/engine/ir/datasetops/source/usps_node.h"
@@ -248,6 +249,17 @@ PYBIND_REGISTER(MnistNode, 2, ([](const py::module *m) {
                     }));
                 }));
 
+PYBIND_REGISTER(QMnistNode, 2, ([](const py::module *m) {
+                  (void)py::class_<QMnistNode, DatasetNode, std::shared_ptr<QMnistNode>>(*m, "QMnistNode",
+                                                                                         "to create an QMnistNode")
+                    .def(py::init([](std::string dataset_dir, std::string usage, bool compat, py::handle sampler) {
+                      auto qmnist =
+                        std::make_shared<QMnistNode>(dataset_dir, usage, compat, toSamplerObj(sampler), nullptr);
+                      THROW_IF_ERROR(qmnist->ValidateParams());
+                      return qmnist;
+                    }));
+                }));
+
 PYBIND_REGISTER(RandomNode, 2, ([](const py::module *m) {
                   (void)py::class_<RandomNode, DatasetNode, std::shared_ptr<RandomNode>>(*m, "RandomNode",
                                                                                          "to create a RandomNode")

mindspore/ccsrc/minddata/dataset/engine/datasetops/source/CMakeLists.txt

@@ -21,6 +21,7 @@ set(DATASET_ENGINE_DATASETOPS_SOURCE_SRC_FILES
     cityscapes_op.cc
     div2k_op.cc
     flickr_op.cc
+    qmnist_op.cc
     )
 
 set(DATASET_ENGINE_DATASETOPS_SOURCE_SRC_FILES

mindspore/ccsrc/minddata/dataset/engine/datasetops/source/mnist_op.h

@@ -81,7 +81,7 @@ class MnistOp : public MappableLeafOp {
   // \return DatasetName of the current Op
   virtual std::string DatasetName(bool upper = false) const { return upper ? "Mnist" : "mnist"; }
 
- private:
+ protected:
   // Load a tensor row according to a pair
   // @param row_id_type row_id - id for this tensor row
   // @param ImageLabelPair pair - <imagefile,label>
@@ -94,14 +94,14 @@ class MnistOp : public MappableLeafOp {
   // @param std::ifstream *image_reader - image file stream
   // @param uint32_t num_images - returns the number of images
   // @return Status The status code returned
-  Status CheckImage(const std::string &file_name, std::ifstream *image_reader, uint32_t *num_images);
+  virtual Status CheckImage(const std::string &file_name, std::ifstream *image_reader, uint32_t *num_images);
 
   // Check label stream.
   // @param const std::string &file_name - label file name
   // @param std::ifstream *label_reader - label file stream
   // @param uint32_t num_labels - returns the number of labels
   // @return Status The status code returned
-  Status CheckLabel(const std::string &file_name, std::ifstream *label_reader, uint32_t *num_labels);
+  virtual Status CheckLabel(const std::string &file_name, std::ifstream *label_reader, uint32_t *num_labels);
 
   // Read 4 bytes of data from a file stream.
   // @param std::ifstream *reader - file stream to read
@@ -118,7 +118,7 @@ class MnistOp : public MappableLeafOp {
   // @param std::ifstream *label_reader - label file stream
   // @param int64_t read_num - number of image to read
   // @return Status The status code returned
-  Status ReadImageAndLabel(std::ifstream *image_reader, std::ifstream *label_reader, size_t index);
+  virtual Status ReadImageAndLabel(std::ifstream *image_reader, std::ifstream *label_reader, size_t index);
 
   // Parse all mnist dataset files
   // @return Status The status code returned
@@ -126,7 +126,7 @@ class MnistOp : public MappableLeafOp {
 
   // Read all files in the directory
   // @return Status The status code returned
-  Status WalkAllFiles();
+  virtual Status WalkAllFiles();
 
   // Called first when function is called
   // @return Status The status code returned

mindspore/ccsrc/minddata/dataset/engine/datasetops/source/qmnist_op.cc

@@ -0,0 +1,283 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+#include "minddata/dataset/engine/datasetops/source/qmnist_op.h"
+
+#include <algorithm>
+#include <fstream>
+#include <iomanip>
+#include <set>
+#include <utility>
+
+#include "debug/common.h"
+#include "minddata/dataset/core/config_manager.h"
+#include "minddata/dataset/core/tensor_shape.h"
+#include "minddata/dataset/engine/datasetops/source/sampler/sequential_sampler.h"
+#include "minddata/dataset/engine/db_connector.h"
+#include "minddata/dataset/engine/execution_tree.h"
+#include "utils/file_utils.h"
+#include "utils/ms_utils.h"
+
+namespace mindspore {
+namespace dataset {
+const int32_t kQMnistLabelFileMagicNumber = 3074;
+const int32_t kQMnistImageRows = 28;
+const int32_t kQMnistImageCols = 28;
+const int32_t kQMnistLabelLength = 8;
+
+QMnistOp::QMnistOp(const std::string &folder_path, const std::string &usage, bool compat,
+                   std::unique_ptr<DataSchema> data_schema, std::shared_ptr<SamplerRT> sampler, int32_t num_workers,
+                   int32_t queue_size)
+    : MnistOp(usage, num_workers, folder_path, queue_size, std::move(data_schema), std::move(sampler)),
+      compat_(compat) {}
+
+void QMnistOp::Print(std::ostream &out, bool show_all) const {
+  if (!show_all) {
+    // Call the super class for displaying any common 1-liner info
+    ParallelOp::Print(out, show_all);
+    // Then show any custom derived-internal 1-liner info for this op
+    out << "\n";
+  } else {
+    // Call the super class for displaying any common detailed info
+    ParallelOp::Print(out, show_all);
+    // Then show any custom derived-internal stuff
+    out << "\nNumber of rows: " << num_rows_ << "\n"
+        << DatasetName(true) << " directory: " << folder_path_ << "\nUsage: " << usage_
+        << "\nCompat: " << (compat_ ? "yes" : "no") << "\n\n";
+  }
+}
+
+// Load 1 TensorRow (image, label) using 1 MnistLabelPair or QMnistImageInfoPair.
+Status QMnistOp::LoadTensorRow(row_id_type row_id, TensorRow *trow) {
+  RETURN_UNEXPECTED_IF_NULL(trow);
+  std::shared_ptr<Tensor> image, label;
+  if (compat_) {
+    MnistLabelPair qmnist_pair = image_label_pairs_[row_id];
+    RETURN_IF_NOT_OK(Tensor::CreateFromTensor(qmnist_pair.first, &image));
+    RETURN_IF_NOT_OK(Tensor::CreateScalar(qmnist_pair.second, &label));
+  } else {
+    QMnistImageInfoPair qmnist_pair = image_info_pairs_[row_id];
+    RETURN_IF_NOT_OK(Tensor::CreateFromTensor(qmnist_pair.first, &image));
+    RETURN_IF_NOT_OK(Tensor::CreateFromTensor(qmnist_pair.second, &label));
+  }
+  (*trow) = TensorRow(row_id, {std::move(image), std::move(label)});
+  trow->setPath({image_path_[row_id], label_path_[row_id]});
+  return Status::OK();
+}
+
+Status QMnistOp::CountTotalRows(const std::string &dir, const std::string &usage, int64_t *count) {
+  RETURN_UNEXPECTED_IF_NULL(count);
+  *count = 0;
+
+  auto schema = std::make_unique<DataSchema>();
+  RETURN_IF_NOT_OK(schema->AddColumn(ColDescriptor("image", DataType(DataType::DE_UINT8), TensorImpl::kCv, 1)));
+  TensorShape scalar = TensorShape::CreateScalar();
+  RETURN_IF_NOT_OK(
+    schema->AddColumn(ColDescriptor("label", DataType(DataType::DE_UINT32), TensorImpl::kFlexible, 0, &scalar)));
+
+  const int64_t num_samples = 0;
+  const int64_t start_index = 0;
+  auto sampler = std::make_shared<SequentialSamplerRT>(start_index, num_samples);
+
+  std::shared_ptr<ConfigManager> cfg = GlobalContext::config_manager();
+  int32_t num_workers = cfg->num_parallel_workers();
+  int32_t op_connector_size = cfg->op_connector_size();
+
+  // compat does not affect the count result, so set it to true default.
+  auto op =
+    std::make_shared<QMnistOp>(dir, usage, true, std::move(schema), std::move(sampler), num_workers, op_connector_size);
+
+  // the logic of counting the number of samples
+  RETURN_IF_NOT_OK(op->WalkAllFiles());
+  for (size_t i = 0; i < op->image_names_.size(); ++i) {
+    std::ifstream image_reader;
+    image_reader.open(op->image_names_[i], std::ios::binary);
+    std::ifstream label_reader;
+    label_reader.open(op->label_names_[i], std::ios::binary);
+
+    uint32_t num_images;
+    RETURN_IF_NOT_OK(op->CheckImage(op->image_names_[i], &image_reader, &num_images));
+    uint32_t num_labels;
+    RETURN_IF_NOT_OK(op->CheckLabel(op->label_names_[i], &label_reader, &num_labels));
+    CHECK_FAIL_RETURN_UNEXPECTED((num_images == num_labels),
+                                 "Invalid data, num of images is not equal to num of labels.");
+
+    if (usage == "test10k") {
+      // only use the first 10k samples and drop the last 50k samples
+      num_images = 10000;
+      num_labels = 10000;
+    } else if (usage == "test50k") {
+      // only use the last 50k samples and drop the first 10k samples
+      num_images = 50000;
+      num_labels = 50000;
+    }
+
+    *count = *count + num_images;
+
+    // Close the readers
+    image_reader.close();
+    label_reader.close();
+  }
+
+  return Status::OK();
+}
+
+Status QMnistOp::WalkAllFiles() {
+  const std::string image_ext = "images-idx3-ubyte";
+  const std::string label_ext = "labels-idx2-int";
+  const std::string train_prefix = "qmnist-train";
+  const std::string test_prefix = "qmnist-test";
+  const std::string nist_prefix = "xnist";
+
+  auto real_folder_path = FileUtils::GetRealPath(folder_path_.data());
+  CHECK_FAIL_RETURN_UNEXPECTED(real_folder_path.has_value(), "Get real path failed: " + folder_path_);
+  Path root_dir(real_folder_path.value());
+
+  if (usage_ == "train") {
+    image_names_.push_back((root_dir / Path(train_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(train_prefix + "-" + label_ext)).ToString());
+  } else if (usage_ == "test" || usage_ == "test10k" || usage_ == "test50k") {
+    image_names_.push_back((root_dir / Path(test_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(test_prefix + "-" + label_ext)).ToString());
+  } else if (usage_ == "nist") {
+    image_names_.push_back((root_dir / Path(nist_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(nist_prefix + "-" + label_ext)).ToString());
+  } else if (usage_ == "all") {
+    image_names_.push_back((root_dir / Path(train_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(train_prefix + "-" + label_ext)).ToString());
+    image_names_.push_back((root_dir / Path(test_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(test_prefix + "-" + label_ext)).ToString());
+    image_names_.push_back((root_dir / Path(nist_prefix + "-" + image_ext)).ToString());
+    label_names_.push_back((root_dir / Path(nist_prefix + "-" + label_ext)).ToString());
+  }
+
+  CHECK_FAIL_RETURN_UNEXPECTED(image_names_.size() == label_names_.size(),
+                               "Invalid data, num of images does not equal to num of labels.");
+
+  for (size_t i = 0; i < image_names_.size(); i++) {
+    Path file_path(image_names_[i]);
+    CHECK_FAIL_RETURN_UNEXPECTED(file_path.Exists() && !file_path.IsDirectory(),
+                                 "Failed to find " + DatasetName() + " image file: " + file_path.ToString());
+    MS_LOG(INFO) << DatasetName(true) << " operator found image file at " << file_path.ToString() << ".";
+  }
+
+  for (size_t i = 0; i < label_names_.size(); i++) {
+    Path file_path(label_names_[i]);
+    CHECK_FAIL_RETURN_UNEXPECTED(file_path.Exists() && !file_path.IsDirectory(),
+                                 "Failed to find " + DatasetName() + " label file: " + file_path.ToString());
+    MS_LOG(INFO) << DatasetName(true) << " operator found label file at " << file_path.ToString() << ".";
+  }
+
+  return Status::OK();
+}
+
+Status QMnistOp::ReadImageAndLabel(std::ifstream *image_reader, std::ifstream *label_reader, size_t index) {
+  RETURN_UNEXPECTED_IF_NULL(image_reader);
+  RETURN_UNEXPECTED_IF_NULL(label_reader);
+  uint32_t num_images, num_labels;
+  RETURN_IF_NOT_OK(CheckImage(image_names_[index], image_reader, &num_images));
+  RETURN_IF_NOT_OK(CheckLabel(label_names_[index], label_reader, &num_labels));
+  CHECK_FAIL_RETURN_UNEXPECTED((num_images == num_labels),
+                               "Invalid data, num_images is not equal to num_labels. Ensure data file is not damaged.");
+
+  // The image size of the QMNIST dataset is fixed at [28,28]
+  int64_t image_size = kQMnistImageRows * kQMnistImageCols;
+  int64_t label_length = kQMnistLabelLength;
+
+  if (usage_ == "test10k") {
+    // only use the first 10k samples and drop the last 50k samples
+    num_images = 10000;
+    num_labels = 10000;
+  } else if (usage_ == "test50k") {
+    num_images = 50000;
+    num_labels = 50000;
+    // skip the first 10k samples for ifstream reader
+    (void)image_reader->ignore(image_size * 10000);
+    (void)label_reader->ignore(label_length * 10000 * 4);
+  }
+
+  auto images_buf = std::make_unique<char[]>(image_size * num_images);
+  auto labels_buf = std::make_unique<uint32_t[]>(label_length * num_labels);
+  if (images_buf == nullptr || labels_buf == nullptr) {
+    std::string err_msg = "[Internal ERROR] Failed to allocate memory for " + DatasetName() + " buffer.";
+    MS_LOG(ERROR) << err_msg.c_str();
+    RETURN_STATUS_UNEXPECTED(err_msg);
+  }
+  (void)image_reader->read(images_buf.get(), image_size * num_images);
+  if (image_reader->fail()) {
+    RETURN_STATUS_UNEXPECTED("Invalid file, failed to read " + DatasetName() + " image: " + image_names_[index] +
+                             ", size:" + std::to_string(image_size * num_images) +
+                             ". Ensure data file is not damaged.");
+  }
+  // uint32_t use 4 bytes in memory
+  (void)label_reader->read(reinterpret_cast<char *>(labels_buf.get()), label_length * num_labels * 4);
+  if (label_reader->fail()) {
+    RETURN_STATUS_UNEXPECTED("Invalid file, failed to read " + DatasetName() + " label:" + label_names_[index] +
+                             ", size: " + std::to_string(label_length * num_labels) +
+                             ". Ensure data file is not damaged.");
+  }
+  TensorShape image_tensor_shape = TensorShape({kQMnistImageRows, kQMnistImageCols, 1});
+  TensorShape label_tensor_shape = TensorShape({kQMnistLabelLength});
+  for (int64_t data_index = 0; data_index != num_images; data_index++) {
+    auto image = &images_buf[data_index * image_size];
+    for (int64_t image_index = 0; image_index < image_size; image_index++) {
+      image[image_index] = (image[image_index] == 0) ? 0 : 255;
+    }
+    std::shared_ptr<Tensor> image_tensor;
+    RETURN_IF_NOT_OK(Tensor::CreateFromMemory(image_tensor_shape, data_schema_->Column(0).Type(),
+                                              reinterpret_cast<unsigned char *>(image), &image_tensor));
+
+    auto label = &labels_buf[data_index * label_length];
+    for (int64_t label_index = 0; label_index < label_length; label_index++) {
+      label[label_index] = SwapEndian(label[label_index]);
+    }
+    std::shared_ptr<Tensor> label_tensor;
+    RETURN_IF_NOT_OK(Tensor::CreateFromMemory(label_tensor_shape, data_schema_->Column(1).Type(),
+                                              reinterpret_cast<unsigned char *>(label), &label_tensor));
+
+    image_info_pairs_.emplace_back(std::make_pair(image_tensor, label_tensor));
+    image_label_pairs_.emplace_back(std::make_pair(image_tensor, label[0]));
+    image_path_.push_back(image_names_[index]);
+    label_path_.push_back(label_names_[index]);
+  }
+  return Status::OK();
+}
+
+Status QMnistOp::CheckLabel(const std::string &file_name, std::ifstream *label_reader, uint32_t *num_labels) {
+  RETURN_UNEXPECTED_IF_NULL(label_reader);
+  RETURN_UNEXPECTED_IF_NULL(num_labels);
+  CHECK_FAIL_RETURN_UNEXPECTED(label_reader->is_open(),
+                               "Invalid file, failed to open " + DatasetName() + " label file: " + file_name);
+  int64_t label_len = label_reader->seekg(0, std::ios::end).tellg();
+  (void)label_reader->seekg(0, std::ios::beg);
+  // The first 12 bytes of the label file are type, number and length
+  CHECK_FAIL_RETURN_UNEXPECTED(label_len >= 12, "Invalid file, " + DatasetName() + " file is corrupted: " + file_name);
+  uint32_t magic_number;
+  RETURN_IF_NOT_OK(ReadFromReader(label_reader, &magic_number));
+  CHECK_FAIL_RETURN_UNEXPECTED(magic_number == kQMnistLabelFileMagicNumber,
+                               "Invalid file, this is not the " + DatasetName() + " label file: " + file_name);
+  uint32_t num_items;
+  RETURN_IF_NOT_OK(ReadFromReader(label_reader, &num_items));
+  uint32_t length;
+  RETURN_IF_NOT_OK(ReadFromReader(label_reader, &length));
+  CHECK_FAIL_RETURN_UNEXPECTED(length == kQMnistLabelLength, "Invalid data, length of labels is not equal to 8.");
+
+  CHECK_FAIL_RETURN_UNEXPECTED((label_len - 12) == num_items * kQMnistLabelLength * 4,
+                               "Invalid data, number of labels is wrong.");
+  *num_labels = num_items;
+  return Status::OK();
+}
+}  // namespace dataset
+}  // namespace mindspore

mindspore/ccsrc/minddata/dataset/engine/datasetops/source/qmnist_op.h

@@ -0,0 +1,113 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+#ifndef MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_DATASETOPS_SOURCE_QMNIST_OP_H_
+#define MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_DATASETOPS_SOURCE_QMNIST_OP_H_
+
+#include <algorithm>
+#include <map>
+#include <memory>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "minddata/dataset/core/tensor.h"
+#include "minddata/dataset/engine/data_schema.h"
+#include "minddata/dataset/engine/datasetops/parallel_op.h"
+#include "minddata/dataset/engine/datasetops/source/mappable_leaf_op.h"
+#include "minddata/dataset/engine/datasetops/source/mnist_op.h"
+#include "minddata/dataset/engine/datasetops/source/sampler/sampler.h"
+#include "minddata/dataset/util/path.h"
+#include "minddata/dataset/util/queue.h"
+#include "minddata/dataset/util/status.h"
+#include "minddata/dataset/util/wait_post.h"
+
+namespace mindspore {
+namespace dataset {
+
+using QMnistImageInfoPair = std::pair<std::shared_ptr<Tensor>, std::shared_ptr<Tensor>>;
+
+class QMnistOp : public MnistOp {
+ public:
+  // Constructor.
+  // @param const std::string &folder_path - dir directory of QMNIST data file.
+  // @param const std::string &usage - Usage of this dataset, can be 'train', 'test', 'test10k', 'test50k', 'nist' or
+  //     'all'.
+  // @param bool compat - Compatibility with Mnist.
+  // @param std::unique_ptr<DataSchema> data_schema - the schema of the QMNIST dataset.
+  // @param td::unique_ptr<Sampler> sampler - sampler tells QMnistOp what to read.
+  // @param int32_t num_workers - number of workers reading images in parallel.
+  // @param int32_t queue_size - connector queue size.
+  QMnistOp(const std::string &folder_path, const std::string &usage, bool compat,
+           std::unique_ptr<DataSchema> data_schema, std::shared_ptr<SamplerRT> sampler, int32_t num_workers,
+           int32_t queue_size);
+
+  // Destructor.
+  ~QMnistOp() = default;
+
+  // Op name getter.
+  // @return std::string - Name of the current Op.
+  std::string Name() const override { return "QMnistOp"; }
+
+  // DatasetName name getter
+  // \return std::string - DatasetName of the current Op
+  std::string DatasetName(bool upper = false) const { return upper ? "QMnist" : "qmnist"; }
+
+  // A print method typically used for debugging.
+  // @param std::ostream &out - out stream.
+  // @param bool show_all - whether to show all information.
+  void Print(std::ostream &out, bool show_all) const override;
+
+  // Function to count the number of samples in the QMNIST dataset.
+  // @param const std::string &dir - path to the QMNIST directory.
+  // @param const std::string &usage - Usage of this dataset, can be 'train', 'test', 'test10k', 'test50k', 'nist' or
+  //     'all'.
+  // @param int64_t *count - output arg that will hold the actual dataset size.
+  // @return Status -The status code returned.
+  static Status CountTotalRows(const std::string &dir, const std::string &usage, int64_t *count);
+
+ private:
+  // Load a tensor row according to a pair.
+  // @param row_id_type row_id - id for this tensor row.
+  // @param TensorRow row - image & label read into this tensor row.
+  // @return Status - The status code returned.
+  Status LoadTensorRow(row_id_type row_id, TensorRow *row) override;
+
+  // Get needed files in the folder_path_.
+  // @return Status - The status code returned.
+  Status WalkAllFiles() override;
+
+  // Read images and labels from the file stream.
+  // @param std::ifstream *image_reader - image file stream.
+  // @param std::ifstream *label_reader - label file stream.
+  // @param size_t index - the index of file that is reading.
+  // @return Status The status code returned.
+  Status ReadImageAndLabel(std::ifstream *image_reader, std::ifstream *label_reader, size_t index) override;
+
+  // Check label stream.
+  // @param const std::string &file_name - label file name.
+  // @param std::ifstream *label_reader - label file stream.
+  // @param uint32_t num_labels - returns the number of labels.
+  // @return Status The status code returned.
+  Status CheckLabel(const std::string &file_name, std::ifstream *label_reader, uint32_t *num_labels) override;
+
+  const bool compat_;  // compatible with mnist
+
+  std::vector<QMnistImageInfoPair> image_info_pairs_;
+};
+}  // namespace dataset
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_DATASETOPS_SOURCE_QMNIST_OP_H_

mindspore/ccsrc/minddata/dataset/engine/ir/datasetops/dataset_node.h

@@ -89,6 +89,7 @@ constexpr char kImageFolderNode[] = "ImageFolderDataset";
 constexpr char kManifestNode[] = "ManifestDataset";
 constexpr char kMindDataNode[] = "MindDataDataset";
 constexpr char kMnistNode[] = "MnistDataset";
+constexpr char kQMnistNode[] = "QMnistDataset";
 constexpr char kRandomNode[] = "RandomDataset";
 constexpr char kSBUNode[] = "SBUDataset";
 constexpr char kTextFileNode[] = "TextFileDataset";

mindspore/ccsrc/minddata/dataset/engine/ir/datasetops/source/CMakeLists.txt

@@ -17,6 +17,7 @@ set(DATASET_ENGINE_IR_DATASETOPS_SOURCE_SRC_FILES
         manifest_node.cc
         minddata_node.cc
         mnist_node.cc
+        qmnist_node.cc
         random_node.cc
         sbu_node.cc
         text_file_node.cc

mindspore/ccsrc/minddata/dataset/engine/ir/datasetops/source/qmnist_node.cc

@@ -0,0 +1,150 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#include "minddata/dataset/engine/ir/datasetops/source/qmnist_node.h"
+
+#include <memory>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "minddata/dataset/engine/datasetops/source/qmnist_op.h"
+#ifndef ENABLE_ANDROID
+#include "minddata/dataset/engine/serdes.h"
+#endif
+#include "minddata/dataset/util/status.h"
+
+namespace mindspore {
+namespace dataset {
+
+QMnistNode::QMnistNode(const std::string &dataset_dir, const std::string &usage, bool compat,
+                       std::shared_ptr<SamplerObj> sampler, std::shared_ptr<DatasetCache> cache)
+    : MappableSourceNode(std::move(cache)),
+      dataset_dir_(dataset_dir),
+      usage_(usage),
+      compat_(compat),
+      sampler_(sampler) {}
+
+std::shared_ptr<DatasetNode> QMnistNode::Copy() {
+  std::shared_ptr<SamplerObj> sampler = (sampler_ == nullptr) ? nullptr : sampler_->SamplerCopy();
+  auto node = std::make_shared<QMnistNode>(dataset_dir_, usage_, compat_, sampler, cache_);
+  return node;
+}
+
+void QMnistNode::Print(std::ostream &out) const {
+  out << (Name() + "(dataset dir: " + dataset_dir_ + ", usage: " + usage_ +
+          ", compat: " + (compat_ ? "true" : "false") + ", cache: " + ((cache_ != nullptr) ? "true" : "false") + ")");
+}
+
+Status QMnistNode::ValidateParams() {
+  RETURN_IF_NOT_OK(DatasetNode::ValidateParams());
+  RETURN_IF_NOT_OK(ValidateDatasetDirParam("QMnistNode", dataset_dir_));
+  RETURN_IF_NOT_OK(ValidateDatasetSampler("QMnistNode", sampler_));
+  RETURN_IF_NOT_OK(ValidateStringValue("QMnistNode", usage_, {"train", "test", "test10k", "test50k", "nist", "all"}));
+  return Status::OK();
+}
+
+Status QMnistNode::Build(std::vector<std::shared_ptr<DatasetOp>> *const node_ops) {
+  // Do internal Schema generation.
+  auto schema = std::make_unique<DataSchema>();
+  RETURN_IF_NOT_OK(schema->AddColumn(ColDescriptor("image", DataType(DataType::DE_UINT8), TensorImpl::kCv, 1)));
+  if (compat_) {
+    TensorShape scalar = TensorShape::CreateScalar();
+    RETURN_IF_NOT_OK(
+      schema->AddColumn(ColDescriptor("label", DataType(DataType::DE_UINT32), TensorImpl::kFlexible, 0, &scalar)));
+  } else {
+    RETURN_IF_NOT_OK(
+      schema->AddColumn(ColDescriptor("label", DataType(DataType::DE_UINT32), TensorImpl::kFlexible, 1)));
+  }
+
+  std::shared_ptr<SamplerRT> sampler_rt = nullptr;
+  RETURN_IF_NOT_OK(sampler_->SamplerBuild(&sampler_rt));
+
+  auto op = std::make_shared<QMnistOp>(dataset_dir_, usage_, compat_, std::move(schema), std::move(sampler_rt),
+                                       num_workers_, connector_que_size_);
+  op->set_total_repeats(GetTotalRepeats());
+  op->set_num_repeats_per_epoch(GetNumRepeatsPerEpoch());
+  node_ops->push_back(op);
+
+  return Status::OK();
+}
+
+// Get the shard id of node
+Status QMnistNode::GetShardId(int32_t *shard_id) {
+  *shard_id = sampler_->ShardId();
+
+  return Status::OK();
+}
+
+// Get Dataset size
+Status QMnistNode::GetDatasetSize(const std::shared_ptr<DatasetSizeGetter> &size_getter, bool estimate,
+                                  int64_t *dataset_size) {
+  if (dataset_size_ > 0) {
+    *dataset_size = dataset_size_;
+    return Status::OK();
+  }
+  int64_t num_rows, sample_size;
+  RETURN_IF_NOT_OK(QMnistOp::CountTotalRows(dataset_dir_, usage_, &num_rows));
+  std::shared_ptr<SamplerRT> sampler_rt = nullptr;
+  RETURN_IF_NOT_OK(sampler_->SamplerBuild(&sampler_rt));
+  sample_size = sampler_rt->CalculateNumSamples(num_rows);
+  if (sample_size == -1) {
+    RETURN_IF_NOT_OK(size_getter->DryRun(shared_from_this(), &sample_size));
+  }
+  *dataset_size = sample_size;
+  dataset_size_ = *dataset_size;
+  return Status::OK();
+}
+
+Status QMnistNode::to_json(nlohmann::json *out_json) {
+  nlohmann::json args, sampler_args;
+  RETURN_IF_NOT_OK(sampler_->to_json(&sampler_args));
+  args["sampler"] = sampler_args;
+  args["num_parallel_workers"] = num_workers_;
+  args["dataset_dir"] = dataset_dir_;
+  args["usage"] = usage_;
+  args["compat"] = compat_;
+  if (cache_ != nullptr) {
+    nlohmann::json cache_args;
+    RETURN_IF_NOT_OK(cache_->to_json(&cache_args));
+    args["cache"] = cache_args;
+  }
+  *out_json = args;
+  return Status::OK();
+}
+
+#ifndef ENABLE_ANDROID
+Status QMnistNode::from_json(nlohmann::json json_obj, std::shared_ptr<DatasetNode> *ds) {
+  CHECK_FAIL_RETURN_UNEXPECTED(json_obj.find("num_parallel_workers") != json_obj.end(),
+                               "Failed to find num_parallel_workers");
+  CHECK_FAIL_RETURN_UNEXPECTED(json_obj.find("dataset_dir") != json_obj.end(), "Failed to find dataset_dir");
+  CHECK_FAIL_RETURN_UNEXPECTED(json_obj.find("usage") != json_obj.end(), "Failed to find usage");
+  CHECK_FAIL_RETURN_UNEXPECTED(json_obj.find("compat") != json_obj.end(), "Failed to find compat");
+  CHECK_FAIL_RETURN_UNEXPECTED(json_obj.find("sampler") != json_obj.end(), "Failed to find sampler");
+  std::string dataset_dir = json_obj["dataset_dir"];
+  std::string usage = json_obj["usage"];
+  bool compat = json_obj["compat"];
+  std::shared_ptr<SamplerObj> sampler;
+  RETURN_IF_NOT_OK(Serdes::ConstructSampler(json_obj["sampler"], &sampler));
+  std::shared_ptr<DatasetCache> cache = nullptr;
+  RETURN_IF_NOT_OK(DatasetCache::from_json(json_obj, &cache));
+  *ds = std::make_shared<QMnistNode>(dataset_dir, usage, compat, sampler, cache);
+  (*ds)->SetNumWorkers(json_obj["num_parallel_workers"]);
+  return Status::OK();
+}
+#endif
+}  // namespace dataset
+}  // namespace mindspore

mindspore/ccsrc/minddata/dataset/engine/ir/datasetops/source/qmnist_node.h

@@ -0,0 +1,111 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_IR_DATASETOPS_SOURCE_QMNIST_NODE_H_
+#define MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_IR_DATASETOPS_SOURCE_QMNIST_NODE_H_
+
+#include <memory>
+#include <string>
+#include <vector>
+
+#include "minddata/dataset/engine/ir/datasetops/dataset_node.h"
+
+namespace mindspore {
+namespace dataset {
+
+class QMnistNode : public MappableSourceNode {
+ public:
+  /// \brief Constructor.
+  QMnistNode(const std::string &dataset_dir, const std::string &usage, bool compat, std::shared_ptr<SamplerObj> sampler,
+             std::shared_ptr<DatasetCache> cache);
+
+  /// \brief Destructor.
+  ~QMnistNode() = default;
+
+  /// \brief Node name getter.
+  /// \return Name of the current node.
+  std::string Name() const override { return kQMnistNode; }
+
+  /// \brief Print the description.
+  /// \param out - The output stream to write output to.
+  void Print(std::ostream &out) const override;
+
+  /// \brief Copy the node to a new object.
+  /// \return A shared pointer to the new copy.
+  std::shared_ptr<DatasetNode> Copy() override;
+
+  /// \brief a base class override function to create the required runtime dataset op objects for this class.
+  /// \param node_ops - A vector containing shared pointer to the Dataset Ops that this object will create.
+  /// \return Status Status::OK() if build successfully.
+  Status Build(std::vector<std::shared_ptr<DatasetOp>> *const node_ops) override;
+
+  /// \brief Parameters validation.
+  /// \return Status Status::OK() if all the parameters are valid.
+  Status ValidateParams() override;
+
+  /// \brief Get the shard id of node.
+  /// \param[in] shard_id The shard id.
+  /// \return Status Status::OK() if get shard id successfully.
+  Status GetShardId(int32_t *shard_id) override;
+
+  /// \brief Base-class override for GetDatasetSize.
+  /// \param[in] size_getter Shared pointer to DatasetSizeGetter.
+  /// \param[in] estimate This is only supported by some of the ops and it's used to speed up the process of getting
+  ///     dataset size at the expense of accuracy.
+  /// \param[out] dataset_size the size of the dataset.
+  /// \return Status of the function.
+  Status GetDatasetSize(const std::shared_ptr<DatasetSizeGetter> &size_getter, bool estimate,
+                        int64_t *dataset_size) override;
+
+  /// \brief Getter functions.
+  const std::string &DatasetDir() const { return dataset_dir_; }
+
+  /// \brief Getter functions.
+  const std::string &Usage() const { return usage_; }
+
+  /// \brief Getter functions.
+  const bool Compat() const { return compat_; }
+
+  /// \brief Get the arguments of node.
+  /// \param[out] out_json JSON string of all attributes.
+  /// \return Status of the function.
+  Status to_json(nlohmann::json *out_json) override;
+
+#ifndef ENABLE_ANDROID
+  /// \brief Function to read dataset in json
+  /// \param[in] json_obj The JSON object to be deserialized
+  /// \param[out] ds Deserialized dataset
+  /// \return Status The status code returned
+  static Status from_json(nlohmann::json json_obj, std::shared_ptr<DatasetNode> *ds);
+#endif
+
+  /// \brief Sampler getter.
+  /// \return SamplerObj of the current node.
+  std::shared_ptr<SamplerObj> Sampler() override { return sampler_; }
+
+  /// \brief Sampler setter.
+  void SetSampler(std::shared_ptr<SamplerObj> sampler) override { sampler_ = sampler; }
+
+ private:
+  std::string dataset_dir_;
+  std::string usage_;
+  bool compat_;
+  std::shared_ptr<SamplerObj> sampler_;
+};
+
+}  // namespace dataset
+}  // namespace mindspore
+#endif  // MINDSPORE_CCSRC_MINDDATA_DATASET_ENGINE_IR_DATASETOPS_SOURCE_QMNIST_NODE_H_

mindspore/ccsrc/minddata/dataset/include/dataset/datasets.h

@@ -2273,6 +2273,90 @@ inline std::shared_ptr<MnistDataset> Mnist(const std::string &dataset_dir, const
   return std::make_shared<MnistDataset>(StringToChar(dataset_dir), StringToChar(usage), sampler, cache);
 }
 
+/// \class QMnistDataset
+/// \brief A source dataset that reads and parses QMNIST dataset.
+class QMnistDataset : public Dataset {
+ public:
+  /// \brief Constructor of QMnistDataset.
+  /// \param[in] dataset_dir Path to the root directory that contains the dataset.
+  /// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all".
+  /// \param[in] compat Whether the label for each example is class number (compat=true)
+  ///     or the full QMNIST information (compat=false).
+  /// \param[in] sampler Shared pointer to a sampler object used to choose samples from the dataset. If sampler is not
+  ///     given, a `RandomSampler` will be used to randomly iterate the entire dataset.
+  /// \param[in] cache Tensor cache to use.
+  explicit QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                         const std::shared_ptr<Sampler> &sampler, const std::shared_ptr<DatasetCache> &cache);
+
+  /// \brief Constructor of QMnistDataset.
+  /// \param[in] dataset_dir Path to the root directory that contains the dataset.
+  /// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all".
+  /// \param[in] compat Whether the label for each example is class number (compat=true)
+  ///     or the full QMNIST information (compat=false).
+  /// \param[in] sampler Raw pointer to a sampler object used to choose samples from the dataset.
+  /// \param[in] cache Tensor cache to use.
+  explicit QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                         const Sampler *sampler, const std::shared_ptr<DatasetCache> &cache);
+
+  /// \brief Constructor of QMnistDataset.
+  /// \param[in] dataset_dir Path to the root directory that contains the dataset.
+  /// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all".
+  /// \param[in] compat Whether the label for each example is class number (compat=true)
+  ///     or the full QMNIST information (compat=false).
+  /// \param[in] sampler Sampler object used to choose samples from the dataset.
+  /// \param[in] cache Tensor cache to use.
+  explicit QMnistDataset(const std::vector<char> &dataset_dir, const std::vector<char> &usage, bool compat,
+                         const std::reference_wrapper<Sampler> sampler, const std::shared_ptr<DatasetCache> &cache);
+
+  /// Destructor of QMnistDataset.
+  ~QMnistDataset() = default;
+};
+
+/// \brief Function to create a QMnistDataset.
+/// \note The generated dataset has two columns ["image", "label"].
+/// \param[in] dataset_dir Path to the root directory that contains the dataset.
+/// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all" (default = "all").
+/// \param[in] compat Whether the label for each example is class number or the full QMNIST information
+///     (default = true).
+/// \param[in] sampler Shared pointer to a sampler object used to choose samples from the dataset. If sampler is not
+///     given, a `RandomSampler` will be used to randomly iterate the entire dataset (default = RandomSampler()).
+/// \param[in] cache Tensor cache to use (default=nullptr which means no cache is used).
+/// \return Shared pointer to the QMnistDataset.
+inline std::shared_ptr<QMnistDataset> QMnist(
+  const std::string &dataset_dir, const std::string &usage = "all", bool compat = true,
+  const std::shared_ptr<Sampler> &sampler = std::make_shared<RandomSampler>(),
+  const std::shared_ptr<DatasetCache> &cache = nullptr) {
+  return std::make_shared<QMnistDataset>(StringToChar(dataset_dir), StringToChar(usage), compat, sampler, cache);
+}
+
+/// \brief Function to create a QMnistDataset.
+/// \note The generated dataset has two columns ["image", "label"].
+/// \param[in] dataset_dir Path to the root directory that contains the dataset.
+/// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all".
+/// \param[in] compat Whether the label for each example is class number or the full QMNIST information.
+/// \param[in] sampler Raw pointer to a sampler object used to choose samples from the dataset.
+/// \param[in] cache Tensor cache to use (default=nullptr which means no cache is used).
+/// \return Shared pointer to the QMnistDataset.
+inline std::shared_ptr<QMnistDataset> QMnist(const std::string &dataset_dir, const std::string &usage, bool compat,
+                                             const Sampler *sampler,
+                                             const std::shared_ptr<DatasetCache> &cache = nullptr) {
+  return std::make_shared<QMnistDataset>(StringToChar(dataset_dir), StringToChar(usage), compat, sampler, cache);
+}
+
+/// \brief Function to create a QMnistDataset.
+/// \note The generated dataset has two columns ["image", "label"].
+/// \param[in] dataset_dir Path to the root directory that contains the dataset.
+/// \param[in] usage Usage of QMNIST, can be "train", "test", "test10k", "test50k", "nist" or "all".
+/// \param[in] compat Whether the label for each example is class number or the full QMNIST information.
+/// \param[in] sampler Sampler object used to choose samples from the dataset.
+/// \param[in] cache Tensor cache to use (default=nullptr which means no cache is used).
+/// \return Shared pointer to the QMnistDataset.
+inline std::shared_ptr<QMnistDataset> QMnist(const std::string &dataset_dir, const std::string &usage, bool compat,
+                                             const std::reference_wrapper<Sampler> sampler,
+                                             const std::shared_ptr<DatasetCache> &cache = nullptr) {
+  return std::make_shared<QMnistDataset>(StringToChar(dataset_dir), StringToChar(usage), compat, sampler, cache);
+}
+
 /// \brief Function to create a ConcatDataset.
 /// \note Reload "+" operator to concat two datasets.
 /// \param[in] datasets1 Shared pointer to the first dataset to be concatenated.
@@ -2565,15 +2649,14 @@ class USPSDataset : public Dataset {
  public:
   /// \brief Constructor of USPSDataset.
   /// \param[in] dataset_dir Path to the root directory that contains the dataset.
-  /// \param[in] usage Usage of USPS, can be "train", "test" or "all" (Default = "all").
-  /// \param[in] num_samples The number of samples to be included in the dataset
-  ///     (Default = 0 means all samples).
-  /// \param[in] shuffle The mode for shuffling data every epoch (Default=ShuffleMode.kGlobal).
+  /// \param[in] usage Usage of USPS, can be "train", "test" or "all".
+  /// \param[in] num_samples The number of samples to be included in the dataset.
+  /// \param[in] shuffle The mode for shuffling data every epoch.
   ///     Can be any of:
   ///     ShuffleMode.kFalse - No shuffling is performed.
   ///     ShuffleMode.kFiles - Shuffle files only.
   ///     ShuffleMode.kGlobal - Shuffle both the files and samples.
-  /// \param[in] num_shards Number of shards that the dataset should be divided into (Default = 1).
+  /// \param[in] num_shards Number of shards that the dataset should be divided into.
   /// \param[in] shard_id The shard ID within num_shards. This argument should be
   ///     specified only when num_shards is also specified (Default = 0).
   /// \param[in] cache Tensor cache to use (default=nullptr which means no cache is used).

mindspore/ccsrc/minddata/dataset/include/dataset/samplers.h

@@ -44,6 +44,7 @@ class Sampler : std::enable_shared_from_this<Sampler> {
   friend class ManifestDataset;
   friend class MindDataDataset;
   friend class MnistDataset;
+  friend class QMnistDataset;
   friend class RandomDataDataset;
   friend class SBUDataset;
   friend class TextFileDataset;

mindspore/dataset/engine/datasets.py

@@ -66,7 +66,7 @@ from .validators import check_batch, check_shuffle, check_map, check_filter, che
     check_bucket_batch_by_length, check_cluedataset, check_save, check_csvdataset, check_paddeddataset, \
     check_tuple_iterator, check_dict_iterator, check_schema, check_to_device_send, check_flickr_dataset, \
     check_sb_dataset, check_flowers102dataset, check_cityscapes_dataset, check_usps_dataset, check_div2k_dataset, \
-    check_sbu_dataset
+    check_sbu_dataset, check_qmnist_dataset
 from ..core.config import get_callback_timeout, _init_device_info, get_enable_shared_mem, get_num_parallel_workers, \
     get_prefetch_size
 from ..core.datatypes import mstype_to_detype, mstypelist_to_detypelist
@@ -3441,6 +3441,131 @@ class MnistDataset(MappableDataset):
         return cde.MnistNode(self.dataset_dir, self.usage, self.sampler)
 
 
+class QMnistDataset(MappableDataset):
+    """
+    A source dataset for reading and parsing the QMNIST dataset.
+
+    The generated dataset has two columns :py:obj:`[image, label]`.
+    The tensor of column :py:obj:`image` is of the uint8 type.
+    The tensor of column :py:obj:`label` is a scalar when `compat` is True else a tensor both of the uint32 type.
+
+    Args:
+        dataset_dir (str): Path to the root directory that contains the dataset.
+        usage (str, optional): Usage of this dataset, can be `train`, `test`, `test10k`, `test50k`, `nist`
+            or `all` (default=None, will read all samples).
+        compat (bool, optional): Whether the label for each example is class number (compat=True) or the full QMNIST
+            information (compat=False) (default=True).
+        num_samples (int, optional): The number of images to be included in the dataset
+            (default=None, will read all images).
+        num_parallel_workers (int, optional): Number of workers to read the data
+            (default=None, will use value set in the config).
+        shuffle (bool, optional): Whether or not to perform shuffle on the dataset
+            (default=None, expected order behavior shown in the table).
+        sampler (Sampler, optional): Object used to choose samples from the
+            dataset (default=None, expected order behavior shown in the table).
+        num_shards (int, optional): Number of shards that the dataset will be divided into (default=None).
+            When this argument is specified, `num_samples` reflects the maximum sample number of per shard.
+        shard_id (int, optional): The shard ID within `num_shards` (default=None). This
+            argument can only be specified when `num_shards` is also specified.
+        cache (DatasetCache, optional): Use tensor caching service to speed up dataset processing.
+            (default=None, which means no cache is used).
+
+    Raises:
+        RuntimeError: If dataset_dir does not contain data files.
+        RuntimeError: If num_parallel_workers exceeds the max thread numbers.
+        RuntimeError: If sampler and shuffle are specified at the same time.
+        RuntimeError: If sampler and sharding are specified at the same time.
+        RuntimeError: If num_shards is specified but shard_id is None.
+        RuntimeError: If shard_id is specified but num_shards is None.
+        ValueError: If shard_id is invalid (< 0 or >= num_shards).
+
+    Note:
+        - This dataset can take in a `sampler`. `sampler` and `shuffle` are mutually exclusive.
+          The table below shows what input arguments are allowed and their expected behavior.
+
+    .. list-table:: Expected Order Behavior of Using `sampler` and `shuffle`
+       :widths: 25 25 50
+       :header-rows: 1
+
+       * - Parameter `sampler`
+         - Parameter `shuffle`
+         - Expected Order Behavior
+       * - None
+         - None
+         - random order
+       * - None
+         - True
+         - random order
+       * - None
+         - False
+         - sequential order
+       * - Sampler object
+         - None
+         - order defined by sampler
+       * - Sampler object
+         - True
+         - not allowed
+       * - Sampler object
+         - False
+         - not allowed
+
+    Examples:
+        >>> qmnist_dataset_dir = "/path/to/qmnist_dataset_directory"
+        >>>
+        >>> # Read 3 samples from QMNIST train dataset
+        >>> dataset = ds.QMnistDataset(dataset_dir=qmnist_dataset_dir, num_samples=3)
+        >>>
+        >>> # Note: In QMNIST dataset, each dictionary has keys "image" and "label"
+
+    About QMNIST dataset:
+
+    The QMNIST dataset was generated from the original data found in the NIST Special Database 19 with the goal to
+    match the MNIST preprocessing as closely as possible.
+    Through an iterative process, researchers tried to generate an additional 50k images of MNIST-like data.
+    They started with a reconstruction process given in the paper and used the Hungarian algorithm to find the best
+    matches between the original MNIST samples and their reconstructed samples.
+
+    Here is the original QMNIST dataset structure.
+    You can unzip the dataset files into this directory structure and read by MindSpore's API.
+
+    .. code-block::
+
+        .
+        └── qmnist_dataset_dir
+             ├── qmnist-train-images-idx3-ubyte
+             ├── qmnist-train-labels-idx2-int
+             ├── qmnist-test-images-idx3-ubyte
+             ├── qmnist-test-labels-idx2-int
+             ├── xnist-images-idx3-ubyte
+             └── xnist-labels-idx2-int
+
+    Citation:
+
+    .. code-block::
+
+        @incollection{qmnist-2019,
+           title = "Cold Case: The Lost MNIST Digits",
+           author = "Chhavi Yadav and L\'{e}on Bottou",\
+           booktitle = {Advances in Neural Information Processing Systems 32},
+           year = {2019},
+           publisher = {Curran Associates, Inc.},
+        }
+    """
+
+    @check_qmnist_dataset
+    def __init__(self, dataset_dir, usage=None, compat=True, num_samples=None, num_parallel_workers=None,
+                 shuffle=None, sampler=None, num_shards=None, shard_id=None, cache=None):
+        super().__init__(num_parallel_workers=num_parallel_workers, sampler=sampler, num_samples=num_samples,
+                         shuffle=shuffle, num_shards=num_shards, shard_id=shard_id, cache=cache)
+
+        self.dataset_dir = dataset_dir
+        self.usage = replace_none(usage, "all")
+        self.compat = compat
+
+    def parse(self, children=None):
+        return cde.QMnistNode(self.dataset_dir, self.usage, self.compat, self.sampler)
+
+
 class MindDataset(MappableDataset):
     """
     A source dataset for reading and parsing MindRecord dataset.

mindspore/dataset/engine/validators.py

@@ -92,6 +92,36 @@ def check_mnist_cifar_dataset(method):
     return new_method
 
 
+def check_qmnist_dataset(method):
+    """A wrapper that wraps a parameter checker around the original Dataset(QMnistDataset)."""
+
+    @wraps(method)
+    def new_method(self, *args, **kwargs):
+        _, param_dict = parse_user_args(method, *args, **kwargs)
+
+        nreq_param_int = ['num_samples', 'num_parallel_workers', 'num_shards', 'shard_id']
+        nreq_param_bool = ['shuffle', 'compat']
+
+        dataset_dir = param_dict.get('dataset_dir')
+        check_dir(dataset_dir)
+
+        usage = param_dict.get('usage')
+        if usage is not None:
+            check_valid_str(usage, ["train", "test", "test10k", "test50k", "nist", "all"], "usage")
+
+        validate_dataset_param_value(nreq_param_int, param_dict, int)
+        validate_dataset_param_value(nreq_param_bool, param_dict, bool)
+
+        check_sampler_shuffle_shard_options(param_dict)
+
+        cache = param_dict.get('cache')
+        check_cache_option(cache)
+
+        return method(self, *args, **kwargs)
+
+    return new_method
+
+
 def check_manifestdataset(method):
     """A wrapper that wraps a parameter checker around the original Dataset(ManifestDataset)."""
 

tests/ut/cpp/dataset/CMakeLists.txt

@@ -28,6 +28,7 @@ SET(DE_UT_SRCS
         c_api_dataset_manifest_test.cc
         c_api_dataset_minddata_test.cc
         c_api_dataset_ops_test.cc
+        c_api_dataset_qmnist_test.cc
         c_api_dataset_randomdata_test.cc
         c_api_dataset_save.cc
         c_api_dataset_sbu_test.cc

tests/ut/cpp/dataset/c_api_dataset_qmnist_test.cc

@@ -0,0 +1,343 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+#include "common/common.h"
+#include "minddata/dataset/include/dataset/datasets.h"
+
+using namespace mindspore::dataset;
+using mindspore::dataset::DataType;
+using mindspore::dataset::Tensor;
+using mindspore::dataset::TensorShape;
+
+class MindDataTestPipeline : public UT::DatasetOpTesting {
+ protected:
+};
+
+TEST_F(MindDataTestPipeline, TestQMnistTrainDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistTrainDataset.";
+
+  // Create a QMNIST Train Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "train", true, std::make_shared<RandomSampler>(false, 5));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 5);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistTestDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistTestDataset.";
+
+  // Create a QMNIST Test Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "test", true, std::make_shared<RandomSampler>(false, 5));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 5);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistNistDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistNistDataset.";
+
+  // Create a QMNIST Nist Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "nist", true, std::make_shared<RandomSampler>(false, 5));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 5);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistAllDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistAllDataset.";
+
+  // Create a QMNIST All Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "all", true, std::make_shared<RandomSampler>(false, 20));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 20);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistCompatDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistCompatDataset.";
+
+  // Create a QMNIST All Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "all", false, std::make_shared<RandomSampler>(false, 20));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    auto label = row["label"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    MS_LOG(INFO) << "Tensor label shape: " << label.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 20);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistDatasetWithPipeline) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistTrainDatasetWithPipeline.";
+
+  // Create two QMNIST Train Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds1 = QMnist(folder_path, "train", true, std::make_shared<RandomSampler>(false, 5));
+  std::shared_ptr<Dataset> ds2 = QMnist(folder_path, "train", true, std::make_shared<RandomSampler>(false, 5));
+  EXPECT_NE(ds1, nullptr);
+  EXPECT_NE(ds2, nullptr);
+
+  // Create two Repeat operation on ds
+  int32_t repeat_num = 1;
+  ds1 = ds1->Repeat(repeat_num);
+  EXPECT_NE(ds1, nullptr);
+  repeat_num = 1;
+  ds2 = ds2->Repeat(repeat_num);
+  EXPECT_NE(ds2, nullptr);
+
+  // Create two Project operation on ds
+  std::vector<std::string> column_project = {"image", "label"};
+  ds1 = ds1->Project(column_project);
+  EXPECT_NE(ds1, nullptr);
+  ds2 = ds2->Project(column_project);
+  EXPECT_NE(ds2, nullptr);
+
+  // Create a Concat operation on the ds
+  ds1 = ds1->Concat({ds2});
+  EXPECT_NE(ds1, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds1->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, mindspore::MSTensor> row;
+  ASSERT_OK(iter->GetNextRow(&row));
+
+  EXPECT_NE(row.find("image"), row.end());
+  EXPECT_NE(row.find("label"), row.end());
+
+  uint64_t i = 0;
+  while (row.size() != 0) {
+    i++;
+    auto image = row["image"];
+    MS_LOG(INFO) << "Tensor image shape: " << image.Shape();
+    ASSERT_OK(iter->GetNextRow(&row));
+  }
+
+  EXPECT_EQ(i, 10);
+
+  // Manually terminate the pipeline
+  iter->Stop();
+}
+
+TEST_F(MindDataTestPipeline, TestGetQMnistDatasetSize) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestGetQMnistTrainDatasetSize.";
+
+  // Create a QMNIST Train Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "train", true);
+  EXPECT_NE(ds, nullptr);
+
+  EXPECT_EQ(ds->GetDatasetSize(), 10);
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistDatasetGetters) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistTrainDatasetGetters.";
+
+  // Create a QMNIST Train Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "train", true);
+  EXPECT_NE(ds, nullptr);
+
+  EXPECT_EQ(ds->GetDatasetSize(), 10);
+  std::vector<DataType> types = ToDETypes(ds->GetOutputTypes());
+  std::vector<TensorShape> shapes = ToTensorShapeVec(ds->GetOutputShapes());
+  std::vector<std::string> column_names = {"image", "label"};
+  int64_t num_classes = ds->GetNumClasses();
+  EXPECT_EQ(types.size(), 2);
+  EXPECT_EQ(types[0].ToString(), "uint8");
+  EXPECT_EQ(types[1].ToString(), "uint32");
+  EXPECT_EQ(shapes.size(), 2);
+  EXPECT_EQ(shapes[0].ToString(), "<28,28,1>");
+  EXPECT_EQ(shapes[1].ToString(), "<>");
+  EXPECT_EQ(num_classes, -1);
+  EXPECT_EQ(ds->GetBatchSize(), 1);
+  EXPECT_EQ(ds->GetRepeatCount(), 1);
+
+  EXPECT_EQ(ds->GetDatasetSize(), 10);
+  EXPECT_EQ(ToDETypes(ds->GetOutputTypes()), types);
+  EXPECT_EQ(ToTensorShapeVec(ds->GetOutputShapes()), shapes);
+  EXPECT_EQ(ds->GetNumClasses(), -1);
+
+  EXPECT_EQ(ds->GetColumnNames(), column_names);
+  EXPECT_EQ(ds->GetDatasetSize(), 10);
+  EXPECT_EQ(ToDETypes(ds->GetOutputTypes()), types);
+  EXPECT_EQ(ToTensorShapeVec(ds->GetOutputShapes()), shapes);
+  EXPECT_EQ(ds->GetBatchSize(), 1);
+  EXPECT_EQ(ds->GetRepeatCount(), 1);
+  EXPECT_EQ(ds->GetNumClasses(), -1);
+  EXPECT_EQ(ds->GetDatasetSize(), 10);
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistDataFail) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistDataFail.";
+
+  // Create a QMNIST Dataset
+  std::shared_ptr<Dataset> ds = QMnist("", "train", true, std::make_shared<RandomSampler>(false, 5));
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  // Expect failure: invalid QMNIST input
+  EXPECT_EQ(iter, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistDataWithInvalidUsageFail) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistDataWithInvalidUsageFail.";
+
+  // Create a QMNIST Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "validation", true);
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  // Expect failure: invalid QMNIST input, validation is not a valid usage
+  EXPECT_EQ(iter, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestQMnistDataWithNullSamplerFail) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestQMnistDataWithNullSamplerFail.";
+
+  // Create a QMNIST Dataset
+  std::string folder_path = datasets_root_path_ + "/testQMnistData/";
+  std::shared_ptr<Dataset> ds = QMnist(folder_path, "train", true, nullptr);
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  // Expect failure: invalid QMNIST input, sampler cannot be nullptr
+  EXPECT_EQ(iter, nullptr);
+}

tests/ut/python/dataset/test_datasets_qmnist.py

@@ -0,0 +1,343 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ==============================================================================
+"""
+Test QMnistDataset operator
+"""
+import os
+
+import matplotlib.pyplot as plt
+import numpy as np
+import pytest
+
+import mindspore.dataset as ds
+import mindspore.dataset.vision.c_transforms as vision
+from mindspore import log as logger
+
+DATA_DIR = "../data/dataset/testQMnistData"
+
+
+def load_qmnist(path, usage, compat=True):
+    """
+    load QMNIST data
+    """
+    image_path = []
+    label_path = []
+    image_ext = "images-idx3-ubyte"
+    label_ext = "labels-idx2-int"
+    train_prefix = "qmnist-train"
+    test_prefix = "qmnist-test"
+    nist_prefix = "xnist"
+    assert usage in ["train", "test", "nist", "all"]
+    if usage == "train":
+        image_path.append(os.path.realpath(os.path.join(path, train_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, train_prefix + "-" + label_ext)))
+    elif usage == "test":
+        image_path.append(os.path.realpath(os.path.join(path, test_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, test_prefix + "-" + label_ext)))
+    elif usage == "nist":
+        image_path.append(os.path.realpath(os.path.join(path, nist_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, nist_prefix + "-" + label_ext)))
+    elif usage == "all":
+        image_path.append(os.path.realpath(os.path.join(path, train_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, train_prefix + "-" + label_ext)))
+        image_path.append(os.path.realpath(os.path.join(path, test_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, test_prefix + "-" + label_ext)))
+        image_path.append(os.path.realpath(os.path.join(path, nist_prefix + "-" + image_ext)))
+        label_path.append(os.path.realpath(os.path.join(path, nist_prefix + "-" + label_ext)))
+
+    assert len(image_path) == len(label_path)
+
+    images = []
+    labels = []
+    for i, _ in enumerate(image_path):
+        with open(image_path[i], 'rb') as image_file:
+            image_file.read(16)
+            image = np.fromfile(image_file, dtype=np.uint8)
+            image = image.reshape(-1, 28, 28, 1)
+            image[image > 0] = 255  # Perform binarization to maintain consistency with our API
+            images.append(image)
+        with open(label_path[i], 'rb') as label_file:
+            label_file.read(12)
+            label = np.fromfile(label_file, dtype='>u4')
+            label = label.reshape(-1, 8)
+            labels.append(label)
+
+    images = np.concatenate(images, 0)
+    labels = np.concatenate(labels, 0)
+    if compat:
+        return images, labels[:, 0]
+    return images, labels
+
+
+def visualize_dataset(images, labels):
+    """
+    Helper function to visualize the dataset samples
+    """
+    num_samples = len(images)
+    for i in range(num_samples):
+        plt.subplot(1, num_samples, i + 1)
+        plt.imshow(images[i].squeeze(), cmap=plt.cm.gray)
+        plt.title(labels[i])
+    plt.show()
+
+
+def test_qmnist_content_check():
+    """
+    Validate QMnistDataset image readings
+    """
+    logger.info("Test QMnistDataset Op with content check")
+    for usage in ["train", "test", "nist", "all"]:
+        data1 = ds.QMnistDataset(DATA_DIR, usage, True, num_samples=10, shuffle=False)
+        images, labels = load_qmnist(DATA_DIR, usage, True)
+        num_iter = 0
+        # in this example, each dictionary has keys "image" and "label"
+        image_list, label_list = [], []
+        for i, data in enumerate(data1.create_dict_iterator(num_epochs=1, output_numpy=True)):
+            image_list.append(data["image"])
+            label_list.append("label {}".format(data["label"]))
+            np.testing.assert_array_equal(data["image"], images[i])
+            np.testing.assert_array_equal(data["label"], labels[i])
+            num_iter += 1
+        assert num_iter == 10
+
+    for usage in ["train", "test", "nist", "all"]:
+        data1 = ds.QMnistDataset(DATA_DIR, usage, False, num_samples=10, shuffle=False)
+        images, labels = load_qmnist(DATA_DIR, usage, False)
+        num_iter = 0
+        # in this example, each dictionary has keys "image" and "label"
+        image_list, label_list = [], []
+        for i, data in enumerate(data1.create_dict_iterator(num_epochs=1, output_numpy=True)):
+            image_list.append(data["image"])
+            label_list.append("label {}".format(data["label"]))
+            np.testing.assert_array_equal(data["image"], images[i])
+            np.testing.assert_array_equal(data["label"], labels[i])
+            num_iter += 1
+        assert num_iter == 10
+
+
+def test_qmnist_basic():
+    """
+    Validate QMnistDataset
+    """
+    logger.info("Test QMnistDataset Op")
+
+    # case 1: test loading whole dataset
+    data1 = ds.QMnistDataset(DATA_DIR, "train", True)
+    num_iter1 = 0
+    for _ in data1.create_dict_iterator(num_epochs=1):
+        num_iter1 += 1
+    assert num_iter1 == 10
+
+    # case 2: test num_samples
+    data2 = ds.QMnistDataset(DATA_DIR, "train", True, num_samples=5)
+    num_iter2 = 0
+    for _ in data2.create_dict_iterator(num_epochs=1):
+        num_iter2 += 1
+    assert num_iter2 == 5
+
+    # case 3: test repeat
+    data3 = ds.QMnistDataset(DATA_DIR, "train", True)
+    data3 = data3.repeat(5)
+    num_iter3 = 0
+    for _ in data3.create_dict_iterator(num_epochs=1):
+        num_iter3 += 1
+    assert num_iter3 == 50
+
+    # case 4: test batch with drop_remainder=False
+    data4 = ds.QMnistDataset(DATA_DIR, "train", True, num_samples=10)
+    assert data4.get_dataset_size() == 10
+    assert data4.get_batch_size() == 1
+    data4 = data4.batch(batch_size=7)  # drop_remainder is default to be False
+    assert data4.get_dataset_size() == 2
+    assert data4.get_batch_size() == 7
+    num_iter4 = 0
+    for _ in data4.create_dict_iterator(num_epochs=1):
+        num_iter4 += 1
+    assert num_iter4 == 2
+
+    # case 5: test batch with drop_remainder=True
+    data5 = ds.QMnistDataset(DATA_DIR, "train", True, num_samples=10)
+    assert data5.get_dataset_size() == 10
+    assert data5.get_batch_size() == 1
+    data5 = data5.batch(batch_size=3, drop_remainder=True)  # the rest of incomplete batch will be dropped
+    assert data5.get_dataset_size() == 3
+    assert data5.get_batch_size() == 3
+    num_iter5 = 0
+    for _ in data5.create_dict_iterator(num_epochs=1):
+        num_iter5 += 1
+    assert num_iter5 == 3
+
+    # case 6: test get_col_names
+    dataset = ds.QMnistDataset(DATA_DIR, "train", True, num_samples=10)
+    assert dataset.get_col_names() == ["image", "label"]
+
+
+def test_qmnist_pk_sampler():
+    """
+    Test QMnistDataset with PKSampler
+    """
+    logger.info("Test QMnistDataset Op with PKSampler")
+    golden = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+    sampler = ds.PKSampler(10)
+    data = ds.QMnistDataset(DATA_DIR, "nist", True, sampler=sampler)
+    num_iter = 0
+    label_list = []
+    for item in data.create_dict_iterator(num_epochs=1, output_numpy=True):
+        label_list.append(item["label"])
+        num_iter += 1
+    np.testing.assert_array_equal(golden, label_list)
+    assert num_iter == 10
+
+
+def test_qmnist_sequential_sampler():
+    """
+    Test QMnistDataset with SequentialSampler
+    """
+    logger.info("Test QMnistDataset Op with SequentialSampler")
+    num_samples = 10
+    sampler = ds.SequentialSampler(num_samples=num_samples)
+    data1 = ds.QMnistDataset(DATA_DIR, "train", True, sampler=sampler)
+    data2 = ds.QMnistDataset(DATA_DIR, "train", True, shuffle=False, num_samples=num_samples)
+    label_list1, label_list2 = [], []
+    num_iter = 0
+    for item1, item2 in zip(data1.create_dict_iterator(num_epochs=1), data2.create_dict_iterator(num_epochs=1)):
+        label_list1.append(item1["label"].asnumpy())
+        label_list2.append(item2["label"].asnumpy())
+        num_iter += 1
+    np.testing.assert_array_equal(label_list1, label_list2)
+    assert num_iter == num_samples
+
+
+def test_qmnist_exception():
+    """
+    Test error cases for QMnistDataset
+    """
+    logger.info("Test error cases for MnistDataset")
+    error_msg_1 = "sampler and shuffle cannot be specified at the same time"
+    with pytest.raises(RuntimeError, match=error_msg_1):
+        ds.QMnistDataset(DATA_DIR, "train", True, shuffle=False, sampler=ds.PKSampler(3))
+
+    error_msg_2 = "sampler and sharding cannot be specified at the same time"
+    with pytest.raises(RuntimeError, match=error_msg_2):
+        ds.QMnistDataset(DATA_DIR, "nist", True, sampler=ds.PKSampler(1), num_shards=2, shard_id=0)
+
+    error_msg_3 = "num_shards is specified and currently requires shard_id as well"
+    with pytest.raises(RuntimeError, match=error_msg_3):
+        ds.QMnistDataset(DATA_DIR, "train", True, num_shards=10)
+
+    error_msg_4 = "shard_id is specified but num_shards is not"
+    with pytest.raises(RuntimeError, match=error_msg_4):
+        ds.QMnistDataset(DATA_DIR, "train", True, shard_id=0)
+
+    error_msg_5 = "Input shard_id is not within the required interval"
+    with pytest.raises(ValueError, match=error_msg_5):
+        ds.QMnistDataset(DATA_DIR, "train", True, num_shards=5, shard_id=-1)
+    with pytest.raises(ValueError, match=error_msg_5):
+        ds.QMnistDataset(DATA_DIR, "train", True, num_shards=5, shard_id=5)
+    with pytest.raises(ValueError, match=error_msg_5):
+        ds.QMnistDataset(DATA_DIR, "train", True, num_shards=2, shard_id=5)
+
+    error_msg_6 = "num_parallel_workers exceeds"
+    with pytest.raises(ValueError, match=error_msg_6):
+        ds.QMnistDataset(DATA_DIR, "train", True, shuffle=False, num_parallel_workers=0)
+    with pytest.raises(ValueError, match=error_msg_6):
+        ds.QMnistDataset(DATA_DIR, "train", True, shuffle=False, num_parallel_workers=256)
+    with pytest.raises(ValueError, match=error_msg_6):
+        ds.QMnistDataset(DATA_DIR, "train", True, shuffle=False, num_parallel_workers=-2)
+
+    error_msg_7 = "Argument shard_id"
+    with pytest.raises(TypeError, match=error_msg_7):
+        ds.QMnistDataset(DATA_DIR, "train", True, num_shards=2, shard_id="0")
+
+    def exception_func(item):
+        raise Exception("Error occur!")
+
+    error_msg_8 = "The corresponding data files"
+    with pytest.raises(RuntimeError, match=error_msg_8):
+        data = ds.QMnistDataset(DATA_DIR, "train", True)
+        data = data.map(operations=exception_func, input_columns=["image"], num_parallel_workers=1)
+        for _ in data.__iter__():
+            pass
+    with pytest.raises(RuntimeError, match=error_msg_8):
+        data = ds.QMnistDataset(DATA_DIR, "train", True)
+        data = data.map(operations=vision.Decode(), input_columns=["image"], num_parallel_workers=1)
+        data = data.map(operations=exception_func, input_columns=["image"], num_parallel_workers=1)
+        for _ in data.__iter__():
+            pass
+    with pytest.raises(RuntimeError, match=error_msg_8):
+        data = ds.QMnistDataset(DATA_DIR, "train", True)
+        data = data.map(operations=exception_func, input_columns=["label"], num_parallel_workers=1)
+        for _ in data.__iter__():
+            pass
+
+
+def test_qmnist_visualize(plot=False):
+    """
+    Visualize QMnistDataset results
+    """
+    logger.info("Test QMnistDataset visualization")
+
+    data1 = ds.QMnistDataset(DATA_DIR, "train", True, num_samples=10, shuffle=False)
+    num_iter = 0
+    image_list, label_list = [], []
+    for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True):
+        image = item["image"]
+        label = item["label"]
+        image_list.append(image)
+        label_list.append("label {}".format(label))
+        assert isinstance(image, np.ndarray)
+        assert image.shape == (28, 28, 1)
+        assert image.dtype == np.uint8
+        assert label.dtype == np.uint32
+        num_iter += 1
+    assert num_iter == 10
+    if plot:
+        visualize_dataset(image_list, label_list)
+
+
+def test_qmnist_usage():
+    """
+    Validate QMnistDataset image readings
+    """
+    logger.info("Test QMnistDataset usage flag")
+
+    def test_config(usage, path=None):
+        path = DATA_DIR if path is None else path
+        try:
+            data = ds.QMnistDataset(path, usage=usage, compat=True, shuffle=False)
+            num_rows = 0
+            for _ in data.create_dict_iterator(num_epochs=1, output_numpy=True):
+                num_rows += 1
+        except (ValueError, TypeError, RuntimeError) as e:
+            return str(e)
+        return num_rows
+
+    assert test_config("train") == 10
+    assert test_config("test") == 10
+    assert test_config("nist") == 10
+    assert test_config("all") == 30
+    assert "usage is not within the valid set of ['train', 'test', 'test10k', 'test50k', 'nist', 'all']" in\
+           test_config("invalid")
+    assert "Argument usage with value ['list'] is not of type [<class 'str'>]" in test_config(["list"])
+
+
+if __name__ == '__main__':
+    test_qmnist_content_check()
+    test_qmnist_basic()
+    test_qmnist_pk_sampler()
+    test_qmnist_sequential_sampler()
+    test_qmnist_exception()
+    test_qmnist_visualize(plot=True)
+    test_qmnist_usage()