Merge pull request #65 from tqchen/master

rename narray to ndarray

Makefile

@@ -54,14 +54,7 @@ else
 	CFLAGS+= -DMXNET_USE_OPENCV=0
 endif
 
-# setup opencv
-ifeq ($(USE_OPENCV_DECODER),1)
-	CFLAGS+= -DMXNET_USE_OPENCV_DECODER=1
-else
-	CFLAGS+= -DMXNET_USE_OPENCV_DECODER=0
-endif
-
-ifeq ($(USE_OPENMP_ITER), 1)
+ifeq ($(USE_OPENMP), 1)
 	CFLAGS += -fopenmp
 endif
 

example/cifar10/cifar10.py

@@ -132,7 +132,7 @@ def RandomInit(narray):
     in_num = narray.shape[1]
     out_num = narray.shape[0]
     a = np.sqrt(3.0 / (in_num + out_num))
-    tmp = mx.narray.array(np.random.uniform(-a, a, narray.shape))
+    tmp = mx.nd.array(np.random.uniform(-a, a, narray.shape))
     narray[:] = tmp
 
 data = mx.symbol.Variable(name="data")
@@ -161,15 +161,16 @@ def RandomInit(narray):
 batch_size = 128
 data_shape = (batch_size, 3, 28, 28)
 
-in_data = mx.narray.empty(data_shape, mx.gpu())
+in_data = mx.nd.empty(data_shape, mx.gpu())
 executor = loss.simple_bind(mx.gpu(), data = in_data)
-out_narray = executor.heads()[0]
-pred = mx.narray.zeros(out_narray.shape, mx.cpu())
+
+out_narray = executor.outputs[0]
+pred = mx.nd.zeros(out_narray.shape, mx.cpu())
 
 arg_narrays, grad_narrays = executor.list_arguments()
 inputs = dict(zip(loss.list_arguments(), arg_narrays))
-tmp_label = mx.narray.zeros(inputs["sm_label"].shape)
-momentum_narrays = [mx.narray.zeros(item.shape, mx.gpu()) for item in grad_narrays]
+tmp_label = mx.nd.zeros(inputs["sm_label"].shape)
+momentum_narrays = [mx.nd.zeros(item.shape, mx.gpu()) for item in grad_narrays]
 
 block = list(zip(grad_narrays, arg_narrays, momentum_narrays))
 

example/mnist/mlp_gpu.py

@@ -30,28 +30,28 @@ def CalAcc(out, label):
 arg_shapes, out_shapes, aux_shapes = softmax.infer_shape(data=data_shape)
 
 # create GPU NArray for data
-arg_narrays = [mx.narray.zeros(shape, ctx=mx.Context("gpu")) for shape in arg_shapes]
-grad_narrays = [mx.narray.zeros(shape, ctx=mx.Context("gpu")) for shape in arg_shapes]
+arg_narrays = [mx.nd.zeros(shape, ctx=mx.gpu()) for shape in arg_shapes]
+grad_narrays = [mx.nd.zeros(shape, ctx=mx.gpu()) for shape in arg_shapes]
 inputs = dict(zip(args_list, arg_narrays))
 
 # create CPU NArray for result stat
 name2shape = dict(zip(args_list, arg_shapes))
-pred = mx.narray.zeros(out_shapes[0])
+pred = mx.nd.zeros(out_shapes[0])
 
 
 # set random weight
 np.random.seed(0)
 for name, narray in inputs.items():
     if "weight" in name:
-        tmp = mx.narray.array(np.random.uniform(-0.07, 0.07, name2shape[name]))
+        tmp = mx.nd.array(np.random.uniform(-0.07, 0.07, name2shape[name]))
         tmp.copyto(narray)
 
 # bind executer
 # TODO(bing): think of a better bind interface
 executor = softmax.bind(mx.Context('gpu'), arg_narrays, grad_narrays)
 # create gradient NArray
-out_narray = executor.heads()[0]
-grad_narray = mx.narray.zeros(out_narray.shape, ctx=mx.Context("gpu"))
+out_narray = executor.outputs[0]
+grad_narray = mx.nd.zeros(out_narray.shape, ctx=mx.gpu())
 
 
 # update
@@ -77,7 +77,7 @@ def Update(grad, weight):
         label="data/t10k-labels-idx1-ubyte",
         batch_size=batch_size, shuffle=True, flat=True, silent=False)
 
-tmp_label = mx.narray.zeros(name2shape["sm_label"])
+tmp_label = mx.nd.zeros(name2shape["sm_label"])
 
 def test_mlp():
     acc_train = 0.

example/mnist/mlp_multi_gpu.py

@@ -43,18 +43,18 @@ def updater(key, grad, weight):
 np.random.seed(0)
 for idx in sync_indices:
     shape = param_shapes[idx]
-    val = mx.narray.zeros(shape)
+    val = mx.nd.zeros(shape)
     if "weight" in param_names[idx]:
         val[:] = np.random.uniform(-0.07, 0.07, shape)
     mx.kvstore.init(idx, val)
 
 # allocate device's memory
-params = [[mx.narray.zeros(s, d) for s in param_shapes] for d in devs]
-grads = [[mx.narray.zeros(s, d) for s in param_shapes] for d in devs]
+params = [[mx.nd.zeros(s, d) for s in param_shapes] for d in devs]
+grads = [[mx.nd.zeros(s, d) for s in param_shapes] for d in devs]
 
 # create executors for devices
 executors = [mlp.bind(devs[d], params[d], grads[d]) for d in range(num_devs)]
-forward_out = [mx.narray.zeros(e.heads()[0].shape) for e in executors]
+forward_out = [mx.nd.zeros(e.heads()[0].shape) for e in executors]
 
 # data reader
 get_data.GetMNIST_ubyte()