removes python path insert of tests folder for examples

example/adversary/adversary_generation.ipynb

@@ -28,10 +28,7 @@
     "import matplotlib.pyplot as plt\n",
     "import matplotlib.cm as cm\n",
     "\n",
-    "import os\n",
-    "import sys\n",
-    "sys.path.append(os.path.join(os.getcwd(), \"../../tests/python/common\"))\n",
-    "from get_data import MNISTIterator"
+    "from mxnet.test_utils import get_mnist_iterator"
    ]
   },
   {
@@ -53,7 +50,7 @@
    "source": [
     "dev = mx.cpu()\n",
     "batch_size = 100\n",
-    "train_iter, val_iter = mnist_iterator(batch_size=batch_size, input_shape = (1,28,28))"
+    "train_iter, val_iter = get_mnist_iterator(batch_size=batch_size, input_shape = (1,28,28))"
    ]
   },
   {

example/caffe/data.py

@@ -15,19 +15,14 @@
 # specific language governing permissions and limitations
 # under the License.
 
-import sys
-import os
-# code to automatically download dataset
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-import get_data
 import mxnet as mx
+from mxnet.test_utils import get_mnist_ubyte
 
 def get_iterator(data_shape, use_caffe_data):
     def get_iterator_impl_mnist(args, kv):
         """return train and val iterators for mnist"""
         # download data
-        get_data.GetMNIST_ubyte()
+        get_mnist_ubyte()
         flat = False if len(data_shape) != 1 else True
 
         train           = mx.io.MNISTIter(

example/gluon/data.py

@@ -19,39 +19,11 @@
 """ data iterator for mnist """
 import os
 import random
-import sys
-# code to automatically download dataset
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-import get_data
 import mxnet as mx
+from mxnet.test_utils import get_cifar10
 
-def mnist_iterator(batch_size, input_shape):
-    """return train and val iterators for mnist"""
-    # download data
-    get_data.GetMNIST_ubyte()
-    flat = False if len(input_shape) == 3 else True
-
-    train_dataiter = mx.io.MNISTIter(
-        image="data/train-images-idx3-ubyte",
-        label="data/train-labels-idx1-ubyte",
-        input_shape=input_shape,
-        batch_size=batch_size,
-        shuffle=True,
-        flat=flat)
-
-    val_dataiter = mx.io.MNISTIter(
-        image="data/t10k-images-idx3-ubyte",
-        label="data/t10k-labels-idx1-ubyte",
-        input_shape=input_shape,
-        batch_size=batch_size,
-        flat=flat)
-
-    return (train_dataiter, val_dataiter)
-
-
-def cifar10_iterator(batch_size, data_shape, resize=-1):
-    get_data.GetCifar10()
+def get_cifar10_iterator(batch_size, data_shape, resize=-1, num_parts=1, part_index=0):
+    get_cifar10()
 
     train = mx.io.ImageRecordIter(
         path_imgrec = "data/cifar/train.rec",
@@ -60,7 +32,9 @@ def cifar10_iterator(batch_size, data_shape, resize=-1):
         data_shape  = data_shape,
         batch_size  = batch_size,
         rand_crop   = True,
-        rand_mirror = True)
+        rand_mirror = True,
+        num_parts=num_parts,
+        part_index=part_index)
 
     val = mx.io.ImageRecordIter(
         path_imgrec = "data/cifar/test.rec",
@@ -69,11 +43,14 @@ def cifar10_iterator(batch_size, data_shape, resize=-1):
         rand_crop   = False,
         rand_mirror = False,
         data_shape  = data_shape,
-        batch_size  = batch_size)
+        batch_size  = batch_size,
+        num_parts=num_parts,
+        part_index=part_index)
 
     return train, val
 
-def imagenet_iterator(train_data, val_data, batch_size, data_shape, resize=-1):
+
+def get_imagenet_iterator(train_data, val_data, batch_size, data_shape, resize=-1, num_parts=1, part_index=0):
     train = mx.io.ImageRecordIter(
         path_imgrec             = train_data,
         data_shape              = data_shape,
@@ -96,7 +73,9 @@ def imagenet_iterator(train_data, val_data, batch_size, data_shape, resize=-1):
         max_random_shear_ratio  = 0.1,
         max_random_aspect_ratio = 0.25,
         fill_value              = 127,
-        min_random_scale        = 0.533)
+        min_random_scale        = 0.533,
+        num_parts               = num_parts,
+        part_index              = part_index)
 
     val = mx.io.ImageRecordIter(
         path_imgrec        = val_data,
@@ -109,7 +88,9 @@ def imagenet_iterator(train_data, val_data, batch_size, data_shape, resize=-1):
         std_b              = 57.375,
         preprocess_threads = 32,
         batch_size         = batch_size,
-        resize             = resize)
+        resize             = resize,
+        num_parts          = num_parts,
+        part_index         = part_index)
 
     return train, val
 

example/gluon/image_classification.py

@@ -26,12 +26,13 @@
 from mxnet.gluon import nn
 from mxnet.gluon.model_zoo import vision as models
 from mxnet import autograd as ag
+from mxnet.test_utils import get_mnist_iterator
 
 from data import *
 
 # CLI
 parser = argparse.ArgumentParser(description='Train a model for image classification.')
-parser.add_argument('--dataset', type=str, default='mnist',
+parser.add_argument('--dataset', type=str, default='cifar10',
                     help='dataset to use. options are mnist, cifar10, and dummy.')
 parser.add_argument('--train-data', type=str, default='',
                     help='training record file to use, required for imagenet.')
@@ -92,25 +93,31 @@
 
 net = models.get_model(opt.model, **kwargs)
 
-# get dataset iterators
-if dataset == 'mnist':
-    train_data, val_data = mnist_iterator(batch_size, (1, 32, 32))
-elif dataset == 'cifar10':
-    train_data, val_data = cifar10_iterator(batch_size, (3, 32, 32))
-elif dataset == 'imagenet':
-    if model_name == 'inceptionv3':
-        train_data, val_data = imagenet_iterator(opt.train_data, opt.val_data,
-                                              batch_size, (3, 299, 299))
-    else:
-        train_data, val_data = imagenet_iterator(opt.train_data, opt.val_data,
-                                                 batch_size, (3, 224, 224))
-elif dataset == 'dummy':
-    if model_name == 'inceptionv3':
-        train_data, val_data = dummy_iterator(batch_size, (3, 299, 299))
-    else:
-        train_data, val_data = dummy_iterator(batch_size, (3, 224, 224))
-
-def test(ctx):
+def get_data_iters(dataset, batch_size, num_workers=1, rank=0):
+    # get dataset iterators
+    if dataset == 'mnist':
+        train_data, val_data = get_mnist_iterator(batch_size, (1, 28, 28),
+                                                  num_parts=num_workers, part_index=rank)
+    elif dataset == 'cifar10':
+        train_data, val_data = get_cifar10_iterator(batch_size, (3, 32, 32),
+                                                    num_parts=num_workers, part_index=rank)
+    elif dataset == 'imagenet':
+        if model_name == 'inceptionv3':
+            train_data, val_data = get_imagenet_iterator(opt.train_data, opt.val_data,
+                                                         batch_size, (3, 299, 299),
+                                                         num_parts=num_workers, part_index=rank)
+        else:
+            train_data, val_data = get_imagenet_iterator(opt.train_data, opt.val_data,
+                                                         batch_size, (3, 224, 224),
+                                                         num_parts=num_workers, part_index=rank)
+    elif dataset == 'dummy':
+        if model_name == 'inceptionv3':
+            train_data, val_data = dummy_iterator(batch_size, (3, 299, 299))
+        else:
+            train_data, val_data = dummy_iterator(batch_size, (3, 224, 224))
+    return train_data, val_data
+
+def test(ctx, val_data):
     metric = mx.metric.Accuracy()
     val_data.reset()
     for batch in val_data:
@@ -127,9 +134,11 @@ def train(epochs, ctx):
     if isinstance(ctx, mx.Context):
         ctx = [ctx]
     net.initialize(mx.init.Xavier(magnitude=2), ctx=ctx)
+    kv = mx.kv.create(opt.kvstore)
+    train_data, val_data = get_data_iters(dataset, batch_size, kv.num_workers, kv.rank)
     trainer = gluon.Trainer(net.collect_params(), 'sgd',
                             {'learning_rate': opt.lr, 'wd': opt.wd, 'momentum': opt.momentum},
-                            kvstore = opt.kvstore)
+                            kvstore = kv)
     metric = mx.metric.Accuracy()
     loss = gluon.loss.SoftmaxCrossEntropyLoss()
 
@@ -164,7 +173,7 @@ def train(epochs, ctx):
         name, acc = metric.get()
         logging.info('[Epoch %d] training: %s=%f'%(epoch, name, acc))
         logging.info('[Epoch %d] time cost: %f'%(epoch, time.time()-tic))
-        name, val_acc = test(ctx)
+        name, val_acc = test(ctx, val_data)
         logging.info('[Epoch %d] validation: %s=%f'%(epoch, name, val_acc))
 
     net.save_params('image-classifier-%s-%d.params'%(opt.model, epochs))
@@ -175,10 +184,12 @@ def main():
         out = net(data)
         softmax = mx.sym.SoftmaxOutput(out, name='softmax')
         mod = mx.mod.Module(softmax, context=[mx.gpu(i) for i in range(num_gpus)] if num_gpus > 0 else [mx.cpu()])
+        kv = mx.kv.create(opt.kvstore)
+        train_data, val_data = get_data_iters(dataset, batch_size, kv.num_workers, kv.rank)
         mod.fit(train_data,
                 eval_data = val_data,
                 num_epoch=opt.epochs,
-                kvstore=opt.kvstore,
+                kvstore=kv,
                 batch_end_callback = mx.callback.Speedometer(batch_size, max(1, opt.log_interval)),
                 epoch_end_callback = mx.callback.do_checkpoint('image-classifier-%s'% opt.model),
                 optimizer = 'sgd',

example/multi-task/example_multi_task.py

@@ -16,12 +16,8 @@
 # under the License.
 
 # pylint: skip-file
-import sys
-import os
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-from get_data import MNISTIterator
 import mxnet as mx
+from mxnet.test_utils import get_mnist_iterator
 import numpy as np
 import logging
 import time
@@ -142,7 +138,7 @@ def get_name_value(self):
 lr = 0.01
 
 network = build_network()
-train, val = MNISTIterator(batch_size=batch_size, input_shape = (784,))
+train, val = get_mnist_iterator(batch_size=batch_size, input_shape = (784,))
 train = Multi_mnist_iterator(train)
 val = Multi_mnist_iterator(val)
 

example/numpy-ops/custom_softmax.py

@@ -16,12 +16,8 @@
 # under the License.
 
 # pylint: skip-file
-import sys
-import os
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-from get_data import MNISTIterator
 import mxnet as mx
+from mxnet.test_utils import get_mnist_iterator
 import numpy as np
 import logging
 
@@ -75,7 +71,7 @@ def create_operator(self, ctx, shapes, dtypes):
 
 # data
 
-train, val = MNISTIterator(batch_size=100, input_shape = (784,))
+train, val = get_mnist_iterator(batch_size=100, input_shape = (784,))
 
 # train
 

example/numpy-ops/ndarray_softmax.py

@@ -16,16 +16,11 @@
 # under the License.
 
 # pylint: skip-file
-import os
-import sys
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-from get_data import MNISTIterator
 import mxnet as mx
+from mxnet.test_utils import get_mnist_iterator
 import numpy as np
 import logging
 
-
 class NDArraySoftmax(mx.operator.NDArrayOp):
     def __init__(self):
         super(NDArraySoftmax, self).__init__(False)
@@ -97,7 +92,7 @@ def backward(self, out_grad, in_data, out_data, in_grad):
 
 # data
 
-train, val = MNISTIterator(batch_size=100, input_shape = (784,))
+train, val = get_mnist_iterator(batch_size=100, input_shape = (784,))
 
 # train
 

example/numpy-ops/numpy_softmax.py

@@ -16,12 +16,8 @@
 # under the License.
 
 # pylint: skip-file
-import sys
-import os
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-from get_data import MNISTIterator
 import mxnet as mx
+from mxnet.test_utils import get_mnist_iterator
 import numpy as np
 import logging
 
@@ -70,7 +66,7 @@ def backward(self, out_grad, in_data, out_data, in_grad):
 
 # data
 
-train, val = MNISTIterator(batch_size=100, input_shape = (784,))
+train, val = get_mnist_iterator(batch_size=100, input_shape = (784,))
 
 # train
 

example/python-howto/monitor_weights.py

@@ -16,12 +16,8 @@
 # under the License.
 
 # pylint: skip-file
-import sys
-import os
-curr_path = os.path.dirname(os.path.abspath(os.path.expanduser(__file__)))
-sys.path.append(os.path.join(curr_path, "../../tests/python/common"))
-from get_data import MNISTIterator
 import mxnet as mx
+from mxnet.test_utils import get_mnist_iterator
 import numpy as np
 import logging
 
@@ -35,7 +31,7 @@
 mlp = mx.symbol.SoftmaxOutput(data = fc3, name = 'softmax')
 
 # data
-train, val = MNISTIterator(batch_size=100, input_shape = (784,))
+train, val = get_mnist_iterator(batch_size=100, input_shape = (784,))
 
 # monitor
 def norm_stat(d):