def test_inception(devs, kv_type):
# guarantee the same weight init for each run
mx.random.seed(0)
logging.basicConfig(level=logging.DEBUG)
(train, val) = common.cifar10(batch_size=128, input_shape=(3, 28, 28))
model = mx.model.FeedForward.create(ctx=devs,
symbol=common.inception(),
X=train,
eval_data=val,
kvstore=kv_type,
num_round=10,
learning_rate=0.1,
momentum=0.9,
wd=0.00001,
initializer=mx.init.Uniform(0.07))
return common.accuracy(model, val)
def test_inception(devs, kv_type):
# guarantee the same weight init for each run
mx.random.seed(0)
logging.basicConfig(level=logging.DEBUG)
(train, val) = common.cifar10(batch_size = 128, input_shape=(3,28,28))
model = mx.model.FeedForward.create(
ctx = devs,
symbol = common.inception(),
X = train,
eval_data = val,
kvstore = kv_type,
num_epoch = 10,
learning_rate = 0.1,
momentum = 0.9,
wd = 0.00001,
initializer = mx.init.Uniform(0.07))
return common.accuracy(model, val)
data_iter.reset()
Y = np.concatenate([y[0].asnumpy() for _, y, _, _ in data_iter])
data_iter.reset()
X = np.concatenate([x[0].asnumpy() for x, _, _, _ in data_iter])
assert X.shape[0] == Y.shape[0]
return (X, Y)
def test_iter(data_iter):
X, Y = get_XY(data_iter)
print X.shape, Y.shape
for i in range(4):
A, B = get_XY(data_iter)
assert (A.shape == X.shape)
assert (B.shape == Y.shape)
assert (np.sum(A != X) == 0)
assert (np.sum(B != Y) == 0)
(train, val) = mnist(batch_size=100, input_shape=(784, ))
test_iter(train)
test_iter(val)
(train, val) = mnist(batch_size=100, input_shape=(1, 28, 28))
test_iter(train)
test_iter(val)
(train, val) = cifar10(batch_size=128, input_shape=(3, 28, 28))
test_iter(train)
test_iter(val)
#!/usr/bin/env python
# pylint: skip-file
import common
import mxnet as mx
import logging
mx.random.seed(0)
logging.basicConfig(level=logging.DEBUG)
kv = mx.kvstore.create('dist_async')
(train, val) = common.cifar10(num_parts = kv.num_workers,
part_index = kv.rank,
batch_size = 128,
input_shape=(3,28,28))
devs = [mx.gpu(i) for i in range(2)]
model = mx.model.FeedForward.create(
ctx = devs,
kvstore = kv,
symbol = common.inception(),
X = train,
eval_data = val,
num_epoch = 20,
learning_rate = 0.05,
momentum = 0.9,
wd = 0.00001,
initializer = mx.init.Uniform(0.07))
common.accuracy(model, val)
def get_XY(data_iter):
data_iter.reset()
Y = np.concatenate([y[0].asnumpy() for _, y, _, _ in data_iter])
data_iter.reset()
X = np.concatenate([x[0].asnumpy() for x, _, _, _ in data_iter])
assert X.shape[0] == Y.shape[0]
return (X,Y)
def test_iter(data_iter):
X, Y = get_XY(data_iter)
print X.shape, Y.shape
for i in range(4):
A, B = get_XY(data_iter)
assert(A.shape == X.shape)
assert(B.shape == Y.shape)
assert(np.sum(A != X) == 0)
assert(np.sum(B != Y) == 0)
(train, val) = mnist(batch_size = 100, input_shape = (784,))
test_iter(train)
test_iter(val)
(train, val) = mnist(batch_size = 100, input_shape=(1,28,28))
test_iter(train)
test_iter(val)
(train, val) = cifar10(batch_size = 128, input_shape=(3,28,28))
test_iter(train)
test_iter(val)
