Merge pull request #5 from leonwanghui/model

Add the first version of TinyMS Model API design

.travis.yml

@@ -16,6 +16,7 @@ install:
 script:
   - python -c "import mindspore; print(mindspore.__version__)"
   - python -c "import tinyms; print(tinyms.__version__)"
+  - pytest
 
 services:
   - docker

tests/st/lenet5.py

@@ -0,0 +1,112 @@
+# Copyright 2020-2021 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.
+# ============================================================================
+"""Lenet5 Tutorial
+The sample can be run on CPU, GPU and Ascend 910 AI processor.
+"""
+import os
+import argparse
+
+import tinyms as ts
+from tinyms import context
+from tinyms.data import MnistDataset, download_dataset
+from tinyms.data.transforms import TypeCast
+from tinyms.vision import Inter, Resize, Rescale, HWC2CHW
+from tinyms.model import Model, lenet5
+from tinyms.callbacks import ModelCheckpoint, CheckpointConfig, LossMonitor
+from tinyms.metrics import Accuracy
+from tinyms.optimizers import Momentum
+from tinyms.losses import SoftmaxCrossEntropyWithLogits
+
+
+def create_dataset(data_path, batch_size=32, repeat_size=1,
+                   num_parallel_workers=1):
+    """ create Mnist dataset for train or eval.
+    Args:
+        data_path: Data path
+        batch_size: The number of data records in each group
+        repeat_size: The number of replicated data records
+        num_parallel_workers: The number of parallel workers
+    """
+    # define dataset
+    mnist_ds = MnistDataset(data_path, num_parallel_workers=num_parallel_workers,
+                            shuffle=True)
+
+    # define map operations
+    c_trans = [
+        Resize((32, 32), interpolation=Inter.LINEAR),  # Resize images to (32, 32)
+        Rescale(1 / 0.3081, -1 * 0.1307 / 0.3081),  # normalize images
+        Rescale(1.0 / 255.0, 0.0),  # rescale images
+        HWC2CHW(),  # change shape from (height, width, channel) to (channel, height, width) to fit network
+    ]
+    type_cast_op = TypeCast(ts.int32)  # change data type of label to int32 to fit network
+
+    # apply map operations on images
+    mnist_ds = mnist_ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=num_parallel_workers)
+    mnist_ds = mnist_ds.map(operations=c_trans, input_columns="image", num_parallel_workers=num_parallel_workers)
+    # apply batch operations
+    mnist_ds = mnist_ds.batch(batch_size, drop_remainder=True)
+    # apply repeat operations
+    mnist_ds = mnist_ds.repeat(repeat_size)
+
+    return mnist_ds
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='MindSpore LeNet Example')
+    parser.add_argument('--device_target', type=str, default="CPU", choices=['Ascend', 'GPU', 'CPU'],
+                        help='device where the code will be implemented (default: CPU)')
+    parser.add_argument('--dataset_path', type=str, default=None, help='Mnist dataset path.')
+    parser.add_argument('--do_eval', type=bool, default=False, help='Do eval or not.')
+    parser.add_argument('--epoch_size', type=int, default=1, help='Epoch size.')
+    parser.add_argument('--batch_size', type=int, default=32, help='Batch size.')
+    parser.add_argument('--checkpoint_path', type=str, default=None, help='CheckPoint file path.')
+    args_opt = parser.parse_args()
+    context.set_context(mode=context.GRAPH_MODE, device_target=args_opt.device_target)
+
+    # download mnist dataset
+    if not args_opt.dataset_path:
+        args_opt.dataset_path = download_dataset('mnist')
+    # build the network
+    net = lenet5()
+    model = Model(net)
+    # define the loss function
+    net_loss = SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
+    # define the optimizer
+    net_opt = Momentum(net.trainable_params(), 0.01, 0.9)
+    model.compile(net_loss, net_opt, metrics={"Accuracy": Accuracy()})
+
+    epoch_size = args_opt.epoch_size
+    batch_size = args_opt.batch_size
+    mnist_path = args_opt.dataset_path
+    dataset_sink_mode = not args_opt.device_target == "CPU"
+
+    if args_opt.do_eval:  # as for evaluation, users could use model.eval
+        print("============== Starting Evaluating ==============")
+        # load testing dataset
+        ds_eval = create_dataset(os.path.join(mnist_path, "test"))
+        # load the saved model for evaluation
+        if args_opt.checkpoint_path:
+            model.load_checkpoint(args_opt.checkpoint_path)
+        acc = model.eval(ds_eval, dataset_sink_mode=dataset_sink_mode)
+        print("============== Accuracy:{} ==============".format(acc))
+    else:  # as for train, users could use model.train
+        print("============== Starting Training ==============")
+        # load training dataset
+        ds_train = create_dataset(os.path.join(mnist_path, "train"), batch_size=batch_size)
+        # save the network model and parameters for subsequence fine-tuning
+        ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet", config=CheckpointConfig(
+            save_checkpoint_steps=1875, keep_checkpoint_max=10))
+        model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor()],
+                    dataset_sink_mode=dataset_sink_mode)

tests/st/resnet50.py

@@ -0,0 +1,113 @@
+# Copyright 2020-2021 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.
+# ============================================================================
+"""ResNet50 Tutorial
+The sample can be run on CPU, GPU and Ascend 910 AI processor.
+"""
+import random
+import argparse
+
+import tinyms as ts
+from tinyms import context
+from tinyms.data import Cifar10Dataset, download_dataset
+from tinyms.data.transforms import TypeCast
+from tinyms.vision import RandomCrop, RandomHorizontalFlip, Resize, Rescale, Normalize, HWC2CHW
+from tinyms.model import Model, resnet50
+from tinyms.callbacks import ModelCheckpoint, CheckpointConfig, LossMonitor
+from tinyms.metrics import Accuracy
+from tinyms.optimizers import Momentum
+from tinyms.losses import SoftmaxCrossEntropyWithLogits
+
+random.seed(1)
+
+
+def create_dataset(data_path, batch_size=32, repeat_size=1, num_parallel_workers=1,
+                   training=True):
+    """ create Cifar10 dataset for train or eval.
+    Args:
+        data_path: Data path
+        batch_size: The number of data records in each group
+        repeat_size: The number of replicated data records
+        num_parallel_workers: The number of parallel workers
+    """
+    # define dataset
+    cifar_ds = Cifar10Dataset(data_path, num_parallel_workers=num_parallel_workers,
+                              shuffle=True)
+
+    # define map operations
+    c_trans = []
+    if training:
+        c_trans += [
+            RandomCrop((32, 32), (4, 4, 4, 4)),
+            RandomHorizontalFlip(prob=0.5),
+        ]
+    c_trans += [
+        Resize((224, 224)), Rescale(1.0 / 255.0, 0.0),
+        Normalize([0.4914, 0.4822, 0.4465], [0.2023, 0.1994, 0.2010]),
+        HWC2CHW(),
+    ]
+    type_cast_op = TypeCast(ts.int32)
+
+    # apply map operations on images
+    cifar_ds = cifar_ds.map(operations=type_cast_op, input_columns="label", num_parallel_workers=num_parallel_workers)
+    cifar_ds = cifar_ds.map(operations=c_trans, input_columns="image", num_parallel_workers=num_parallel_workers)
+    # apply batch operations
+    cifar_ds = cifar_ds.batch(batch_size=batch_size, drop_remainder=True)
+    # apply repeat operations
+    cifar_ds = cifar_ds.repeat(repeat_size)
+
+    return cifar_ds
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='Image classification')
+    parser.add_argument('--device_target', type=str, default="CPU", choices=['Ascend', 'GPU', 'CPU'],
+                        help='device where the code will be implemented (default: CPU)')
+    parser.add_argument('--dataset_path', type=str, default=None, help='Cifar10 dataset path.')
+    parser.add_argument('--do_eval', type=bool, default=False, help='Do eval or not.')
+    parser.add_argument('--epoch_size', type=int, default=1, help='Epoch size.')
+    parser.add_argument('--batch_size', type=int, default=32, help='Batch size.')
+    parser.add_argument('--num_classes', type=int, default=10, help='Num classes.')
+    parser.add_argument('--checkpoint_path', type=str, default=None, help='CheckPoint file path.')
+    args_opt = parser.parse_args()
+    context.set_context(mode=context.GRAPH_MODE, device_target=args_opt.device_target)
+
+    # download cifar10 dataset
+    if not args_opt.dataset_path:
+        args_opt.dataset_path = download_dataset('cifar10')
+    # build the network
+    net = resnet50(args_opt.num_classes)
+    model = Model(net)
+    # define the loss function
+    net_loss = SoftmaxCrossEntropyWithLogits(sparse=True, reduction="mean")
+    # define the optimizer
+    net_opt = Momentum(filter(lambda x: x.requires_grad, net.get_parameters()), 0.01, 0.9)
+    model.compile(loss_fn=net_loss, optimizer=net_opt, metrics={"Accuracy": Accuracy()})
+
+    epoch_size = args_opt.epoch_size
+    batch_size = args_opt.batch_size
+    cifar10_path = args_opt.dataset_path
+    dataset_sink_mode = not args_opt.device_target == "CPU"
+    if args_opt.do_eval:  # as for evaluation, users could use model.eval
+        ds_eval = create_dataset(cifar10_path, batch_size=batch_size, training=False)
+        if args_opt.checkpoint_path:
+            model.load_checkpoint(args_opt.checkpoint_path)
+        acc = model.eval(ds_eval, dataset_sink_mode=dataset_sink_mode)
+        print("============== Accuracy:{} ==============".format(acc))
+    else:  # as for train, users could use model.train
+        ds_train = create_dataset(cifar10_path, batch_size=batch_size)
+        ckpoint_cb = ModelCheckpoint(prefix="resnet_cifar10", config=CheckpointConfig(
+            save_checkpoint_steps=ds_train.get_dataset_size(), keep_checkpoint_max=35))
+        model.train(epoch_size, ds_train, callbacks=[ckpoint_cb, LossMonitor()],
+                    dataset_sink_mode=dataset_sink_mode)

tests/ut/data/test_utils.py

@@ -0,0 +1,41 @@
+# 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.
+# ============================================================================
+import os
+import pytest
+
+from tinyms.data import download_dataset
+
+
+@pytest.mark.skip(reason="no way of currently testing this")
+def test_download_dataset_mnist():
+    download_dataset(dataset_name='mnist', local_path='/tmp')
+
+    assert os.path.exists('/tmp/mnist/train')
+    assert os.path.exists('/tmp/mnist/test')
+
+
+@pytest.mark.skip(reason="no way of currently testing this")
+def test_download_dataset_cifar10():
+    download_dataset(dataset_name='cifar10', local_path='/tmp')
+
+    assert os.path.exists('/tmp/cifar10/cifar-10-batches-bin/batches.meta.txt')
+
+
+@pytest.mark.skip(reason="no way of currently testing this")
+def test_download_dataset_cifar100():
+    download_dataset(dataset_name='cifar100', local_path='/tmp')
+
+    assert os.path.exists('/tmp/cifar100/cifar-100-bin/train.bin')
+    assert os.path.exists('/tmp/cifar100/cifar-100-bin/test.bin')

tests/ut/model/test_model.py

@@ -0,0 +1,33 @@
+# 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.
+# ============================================================================
+
+import tinyms as ts
+from tinyms import context, layers
+from tinyms.layers import SequentialLayer
+from tinyms.model import Model
+
+
+def test_model_predict():
+    context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
+
+    net = SequentialLayer([
+        layers.Conv2d(1, 6, 5, pad_mode='valid', weight_init="ones"),
+        layers.ReLU(),
+        layers.MaxPool2d(kernel_size=2, stride=2)
+    ])
+    model = Model(net)
+    model.compile()
+    z = model.predict(ts.ones((1, 1, 28, 28)))
+    print(z.asnumpy())

tinyms/__init__.py

@@ -14,6 +14,9 @@
 # ============================================================================
 """.. TinyMS package."""
 from .version import __version__
+from . import common
+from .common import *
 
 __all__ = []
 __all__.extend(__version__)
+__all__.extend(common.__all__)

tinyms/callbacks.py

@@ -0,0 +1,20 @@
+# 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.
+# ============================================================================
+
+from mindspore.train import callback
+from mindspore.train.callback import *
+
+__all__ = []
+__all__.extend(callback.__all__)

tinyms/common.py

@@ -0,0 +1,22 @@
+# 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.
+# ============================================================================
+from mindspore import common
+from mindspore.common import *
+from mindspore import numpy
+from mindspore.numpy import *
+
+__all__ = []
+__all__.extend(common.__all__)
+__all__.extend(numpy.__all__)

tinyms/context.py

@@ -0,0 +1,20 @@
+# 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.
+# ============================================================================
+
+from mindspore import context
+from mindspore.context import *
+
+__all__ = []
+__all__.extend(context.__all__)