Dev flow.utils.data part3

docs/source/index.rst

@@ -20,6 +20,8 @@ OneFlow API Reference
     linalg
     image
     optim
+    utils
+
 
 
 Indices and tables

docs/source/utils.rst

@@ -0,0 +1,62 @@
+oneflow.utils
+===================================
+Utils
+----------------------------------
+.. currentmodule:: oneflow.utils
+.. automodule:: oneflow.utils.data
+    :members: DataLoader,
+        Dataset,
+        IterableDataset,
+        TensorDataset,
+        ConcatDataset,
+        Subset,
+        random_split,
+        Sampler,
+        SequentialSampler,
+        RandomSampler,
+        SubsetRandomSampler,
+        BatchSampler
+
+.. currentmodule:: oneflow.utils
+.. automodule:: oneflow.utils.data.distributed
+    :members: DistributedSampler
+
+.. currentmodule:: oneflow.utils
+.. automodule:: oneflow.utils.vision.datasets
+    :members: MNIST,
+        FashionMNIST,
+        CIFAR10,
+        CIFAR100,
+        ImageNet,
+        CocoCaptions,
+        CocoDetection,
+        VOCDetection,
+        VOCSegmentation,
+        DatasetFolder,
+        ImageFolder
+
+.. currentmodule:: oneflow.utils
+.. automodule:: oneflow.utils.vision.transforms
+    :members: Compose,
+        ToTensor,
+        PILToTensor,
+        ConvertImageDtype,
+        ToPILImage,
+        Normalize,
+        Resize,
+        Scale,
+        CenterCrop,
+        Pad,
+        Lambda,
+        RandomTransforms,
+        RandomApply,
+        RandomOrder,
+        RandomChoice,
+        RandomCrop,
+        RandomHorizontalFlip,
+        RandomVerticalFlip,
+        RandomResizedCrop,
+        RandomSizedCrop,
+        FiveCrop,
+        TenCrop,
+        InterpolationMode

python/oneflow/test/dataloader/data_utils.py

@@ -0,0 +1,148 @@
+"""
+Copyright 2020 The OneFlow Authors. 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.
+"""
+import os
+import oneflow as flow
+import oneflow.utils.vision.transforms as transforms
+
+
+def load_data_cifar10(
+    batch_size,
+    data_dir="./data-test/cifar10",
+    download=True,
+    transform=None,
+    source_url=None,
+    num_workers=0,
+):
+    cifar10_train = flow.utils.vision.datasets.CIFAR10(
+        root=data_dir,
+        train=True,
+        download=download,
+        transform=transform,
+        source_url=source_url,
+    )
+    cifar10_test = flow.utils.vision.datasets.CIFAR10(
+        root=data_dir,
+        train=False,
+        download=download,
+        transform=transform,
+        source_url=source_url,
+    )
+
+    train_iter = flow.utils.data.DataLoader(
+        cifar10_train, batch_size=batch_size, shuffle=True, num_workers=num_workers
+    )
+    test_iter = flow.utils.data.DataLoader(
+        cifar10_test, batch_size=batch_size, shuffle=False, num_workers=num_workers
+    )
+    return train_iter, test_iter
+
+
+def load_data_mnist(
+    batch_size, resize=None, root="./data/mnist", download=True, source_url=None
+):
+    """Download the MNIST dataset and then load into memory."""
+    root = os.path.expanduser(root)
+    transformer = []
+    if resize:
+        transformer += [transforms.Resize(resize)]
+    transformer += [transforms.ToTensor()]
+    transformer = transforms.Compose(transformer)
+
+    mnist_train = flow.utils.vision.datasets.MNIST(
+        root=root,
+        train=True,
+        transform=transformer,
+        download=download,
+        source_url=source_url,
+    )
+    mnist_test = flow.utils.vision.datasets.MNIST(
+        root=root,
+        train=False,
+        transform=transformer,
+        download=download,
+        source_url=source_url,
+    )
+    train_iter = flow.utils.data.DataLoader(mnist_train, batch_size, shuffle=True)
+    test_iter = flow.utils.data.DataLoader(mnist_test, batch_size, shuffle=False)
+    return train_iter, test_iter
+
+
+def get_fashion_mnist_dataset(
+    resize=None, root="./data-test/fashion-mnist", download=True, source_url=None,
+):
+    root = os.path.expanduser(root)
+    trans = []
+    if resize:
+        trans.append(transforms.Resize(resize))
+    trans.append(transforms.ToTensor())
+    transform = transforms.Compose(trans)
+
+    mnist_train = flow.utils.vision.datasets.FashionMNIST(
+        root=root,
+        train=True,
+        transform=transform,
+        download=download,
+        source_url=source_url,
+    )
+    mnist_test = flow.utils.vision.datasets.FashionMNIST(
+        root=root,
+        train=False,
+        transform=transform,
+        download=download,
+        source_url=source_url,
+    )
+    return mnist_train, mnist_test
+
+
+# reference: http://tangshusen.me/Dive-into-DL-PyTorch/#/chapter03_DL-basics/3.10_mlp-pytorch
+def load_data_fashion_mnist(
+    batch_size,
+    resize=None,
+    root="./data-test/fashion-mnist",
+    download=True,
+    source_url=None,
+    num_workers=0,
+):
+    """Download the Fashion-MNIST dataset and then load into memory."""
+    root = os.path.expanduser(root)
+    trans = []
+    if resize:
+        trans.append(transforms.Resize(resize))
+    trans.append(transforms.ToTensor())
+    transform = transforms.Compose(trans)
+
+    mnist_train = flow.utils.vision.datasets.FashionMNIST(
+        root=root,
+        train=True,
+        transform=transform,
+        download=download,
+        source_url=source_url,
+    )
+    mnist_test = flow.utils.vision.datasets.FashionMNIST(
+        root=root,
+        train=False,
+        transform=transform,
+        download=download,
+        source_url=source_url,
+    )
+
+    train_iter = flow.utils.data.DataLoader(
+        mnist_train, batch_size, shuffle=True, num_workers=num_workers
+    )
+    test_iter = flow.utils.data.DataLoader(
+        mnist_test, batch_size, shuffle=False, num_workers=num_workers
+    )
+    return train_iter, test_iter

python/oneflow/test/dataloader/test_cifar_dataset.py

@@ -20,6 +20,7 @@
 import oneflow as flow
 import oneflow.nn as nn
 import oneflow.optim as optim
+from data_utils import load_data_cifar10
 
 
 classes = (
@@ -81,21 +82,19 @@ def test(test_case):
         os.getenv("ONEFLOW_TEST_CACHE_DIR", "./data-test"), "cifar10"
     )
 
-    trainset = flow.utils.vision.datasets.CIFAR10(
-        root=data_dir,
-        train=True,
+    train_iter, test_iter = load_data_cifar10(
+        batch_size=batch_size,
+        data_dir=data_dir,
         download=True,
         transform=transform,
         source_url="https://oneflow-public.oss-cn-beijing.aliyuncs.com/datasets/cifar/cifar-10-python.tar.gz",
-    )
-    trainloader = flow.utils.data.DataLoader(
-        trainset, batch_size=batch_size, shuffle=False, num_workers=0
+        num_workers=0,
     )
 
     final_loss = 0
     for epoch in range(1, train_epoch + 1):  # loop over the dataset multiple times
         running_loss = 0.0
-        for i, data in enumerate(trainloader, 1):
+        for i, data in enumerate(train_iter, 1):
             # get the inputs; data is a list of [inputs, labels]
             inputs, labels = data
             inputs = inputs.to(dtype=flow.float32, device=device)
@@ -130,10 +129,3 @@ def test_cifar_dataset(test_case):
 
 if __name__ == "__main__":
     unittest.main()
-    # 1 epoch training log
-    # epoch: 1  step:  2000  loss: 2.107
-    # epoch: 1  step:  4000  loss: 1.838
-    # epoch: 1  step:  6000  loss: 1.644
-    # epoch: 1  step:  8000  loss: 1.535
-    # epoch: 1  step: 10000  loss: 1.528
-    # epoch: 1  step: 12000  loss: 1.476

python/oneflow/test/dataloader/test_fashion_mnist_dataset.py

@@ -20,42 +20,7 @@
 import oneflow.unittest
 import oneflow as flow
 import oneflow.nn as nn
-
-
-# reference: http://tangshusen.me/Dive-into-DL-PyTorch/#/chapter03_DL-basics/3.10_mlp-pytorch
-def load_data_fashion_mnist(
-    batch_size, resize=None, root="./data/fashion-mnist", download=True, source_url=None
-):
-    """Download the Fashion-MNIST dataset and then load into memory."""
-    root = os.path.expanduser(root)
-    transformer = []
-    if resize:
-        transformer += [flow.utils.vision.transforms.Resize(resize)]
-    transformer += [flow.utils.vision.transforms.ToTensor()]
-    transformer = flow.utils.vision.transforms.Compose(transformer)
-
-    mnist_train = flow.utils.vision.datasets.FashionMNIST(
-        root=root,
-        train=True,
-        transform=transformer,
-        download=download,
-        source_url=source_url,
-    )
-    mnist_test = flow.utils.vision.datasets.FashionMNIST(
-        root=root,
-        train=False,
-        transform=transformer,
-        download=download,
-        source_url=source_url,
-    )
-    num_workers = 0
-    train_iter = flow.utils.data.DataLoader(
-        mnist_train, batch_size, shuffle=True, num_workers=num_workers
-    )
-    test_iter = flow.utils.data.DataLoader(
-        mnist_test, batch_size, shuffle=False, num_workers=num_workers
-    )
-    return train_iter, test_iter
+from data_utils import load_data_fashion_mnist
 
 
 def get_fashion_mnist_labels(labels):
@@ -124,7 +89,7 @@ def test(test_case):
     )
     source_url = "https://oneflow-public.oss-cn-beijing.aliyuncs.com/datasets/mnist/Fashion-MNIST/"
     train_iter, test_iter = load_data_fashion_mnist(
-        batch_size, root=data_dir, download=True, source_url=source_url
+        batch_size, resize=None, root=data_dir, download=True, source_url=source_url
     )
     loss = nn.CrossEntropyLoss()
     loss.to(device)
@@ -174,6 +139,3 @@ def test_fashion_mnist_dataset(test_case):
 
 if __name__ == "__main__":
     unittest.main()
-    # 1 epoch training log
-    # epoch 1, loss 0.0034, train acc 0.718, test acc 0.771, cost >>>>>>> 158.32699990272522(s)
-    # epoch 2, loss 0.0022, train acc 0.807, test acc 0.726, cost >>>>>>> 159.64465260505676(s)

python/oneflow/test/dataloader/test_lenet.py

@@ -20,6 +20,7 @@
 import oneflow as flow
 import oneflow.nn as nn
 import oneflow.unittest
+from data_utils import load_data_fashion_mnist
 
 
 # reference: http://tangshusen.me/Dive-into-DL-PyTorch/#/chapter05_CNN/5.5_lenet
@@ -49,46 +50,6 @@ def forward(self, img):
         return output
 
 
-def load_data_fashion_mnist(
-    batch_size,
-    resize=None,
-    root="./data-test/fashion-mnist",
-    download=True,
-    source_url=None,
-    num_workers=0,
-):
-    """Download the Fashion-MNIST dataset and then load into memory."""
-    root = os.path.expanduser(root)
-    trans = []
-    if resize:
-        trans.append(flow.utils.vision.transforms.Resize(resize))
-    trans.append(flow.utils.vision.transforms.ToTensor())
-    transform = flow.utils.vision.transforms.Compose(trans)
-
-    mnist_train = flow.utils.vision.datasets.FashionMNIST(
-        root=root,
-        train=True,
-        transform=transform,
-        download=download,
-        source_url=source_url,
-    )
-    mnist_test = flow.utils.vision.datasets.FashionMNIST(
-        root=root,
-        train=False,
-        transform=transform,
-        download=download,
-        source_url=source_url,
-    )
-
-    train_iter = flow.utils.data.DataLoader(
-        mnist_train, batch_size, shuffle=True, num_workers=num_workers
-    )
-    test_iter = flow.utils.data.DataLoader(
-        mnist_test, batch_size, shuffle=False, num_workers=num_workers
-    )
-    return train_iter, test_iter
-
-
 def evaluate_accuracy(data_iter, net, device=None):
     if device is None and isinstance(net, nn.Module):
         device = list(net.parameters())[0].device
@@ -176,8 +137,3 @@ def test_lenet(test_case):
 
 if __name__ == "__main__":
     unittest.main()
-    # 1 epoch training log
-    # epoch 1, loss 1.1473, train acc 0.569, test acc 0.742, time 162.4 sec
-    # epoch 2, loss 0.5736, train acc 0.784, test acc 0.796, time 158.1 sec
-    # epoch 3, loss 0.4761, train acc 0.826, test acc 0.821, time 154.0 sec
-    # epoch 4, loss 0.4215, train acc 0.848, test acc 0.855, time 160.3 sec