def test_AvgPool():
X = ad.Variable(name="X")
y = ad.avg_pool2d_op(X, kernel_H=2, kernel_W=2, padding=0, stride=2)
executor = ad.Executor([y], ctx=ctx)
X_val = rand.normal(scale=0.1,
size=(batch_size, 10, 10, 10)).astype(np.float32)
res = executor.run(feed_dict={X: X_val})
Check(executor, res, [X], [y], [X_val])
print(sys._getframe().f_code.co_name, 'pass!')
def cnn(executor_ctx=None, num_epochs=10, print_loss_val_each_epoch=False):
print("Build CNN model...")
W1 = init.random_normal((32, 1, 5, 5), stddev=0.1, name='W1')
W2 = init.random_normal((64, 32, 5, 5), stddev=0.1, name='W2')
W3 = init.random_normal((7 * 7 * 64, 10), stddev=0.1, name='W3')
b3 = init.random_normal((10, ), stddev=0.1, name='b3')
X = ad.Variable(name="X")
z1 = ad.conv2d_op(X, W1, padding=2, stride=1)
z2 = ad.relu_op(z1)
z3 = ad.avg_pool2d_op(z2, kernel_H=2, kernel_W=2, padding=0, stride=2)
z4 = ad.conv2d_op(z3, W2, padding=2, stride=1)
z5 = ad.relu_op(z4)
z6 = ad.avg_pool2d_op(z5, kernel_H=2, kernel_W=2, padding=0, stride=2)
z6_flat = ad.array_reshape_op(z6, (-1, 7 * 7 * 64))
y = ad.matmul_op(z6_flat, W3) + b3
executor = ad.Executor([y], ctx=executor_ctx)
rand = np.random.RandomState(seed=123)
X_val = rand.normal(scale=0.1,
size=(batch_size, 1, 28, 28)).astype(np.float32)
ath = executor.run(feed_dict={X: X_val})
hx.hetu2onnx.export(executor, [X], [y], 'ath.onnx')
#
#
sess = rt.InferenceSession("ath.onnx")
input = sess.get_inputs()[0].name
pre = sess.run(None, {input: X_val.astype(np.float32)})[0]
np.testing.assert_allclose(ath[0].asnumpy(), pre, rtol=1e-2)
def version_1(cls, node, tensor_dict, **kwargs):
inputs = [
tensor_dict.get(inp, None) for inp in node.input_tensor_names
]
assert len(inputs) == 1
kernel_shape = node.get_attr_value('kernel_shape')
strides = node.get_attr_value('strides')
assert len(kernel_shape) == 2 and len(strides) == 2
assert strides[0] == strides[1], 'strides 0 and 1 must be equal now!'
#todo,here padding set to 0. check.
y = ad.avg_pool2d_op(inputs[0],
kernel_H=kernel_shape[0],
kernel_W=kernel_shape[1],
padding=0,
stride=strides[0])
tensor_dict[node.output_tensor_names[0]] = y
return y
def conv_relu_avg(x, shape):
weight = init.random_normal(shape=shape, stddev=0.1)
x = ad.conv2d_op(x, weight, padding=2, stride=1)
x = ad.relu_op(x)
x = ad.avg_pool2d_op(x, kernel_H=2, kernel_W=2, padding=0, stride=2)
return x
def resnet_block(x, in_channel, num_blocks, is_first=False, name=''):
if is_first:
out_channel = in_channel
identity = x
x = conv2d(x,
in_channel,
out_channel,
stride=1,
padding=1,
name=name + '_conv1')
x = batch_norm_with_relu(x, out_channel, name + '_bn1')
x = conv2d(x,
out_channel,
out_channel,
stride=1,
padding=1,
name=name + '_conv2')
x = x + identity
else:
out_channel = 2 * in_channel
identity = x
x = batch_norm_with_relu(x, in_channel, name + '_bn0')
x = ad.pad_op(x, [[0, 0], [0, 0], [0, 1], [0, 1]])
x = conv2d(x,
in_channel,
out_channel,
stride=2,
padding=0,
name=name + '_conv1')
x = batch_norm_with_relu(x, out_channel, name + '_bn1')
x = conv2d(x,
out_channel,
out_channel,
stride=1,
padding=1,
name=name + '_conv2')
identity = ad.avg_pool2d_op(identity,
kernel_H=2,
kernel_W=2,
padding=0,
stride=2)
identity = ad.pad_op(
identity,
[[0, 0], [in_channel // 2, in_channel // 2], [0, 0], [0, 0]])
x = x + identity
for i in range(1, num_blocks):
identity = x
x = batch_norm_with_relu(x, out_channel, name + '_bn%d' % (2 * i))
x = conv2d(x,
out_channel,
out_channel,
stride=1,
padding=1,
name=name + '_conv%d' % (2 * i + 1))
x = batch_norm_with_relu(x, out_channel, name + '_bn%d' % (2 * i + 1))
x = conv2d(x,
out_channel,
out_channel,
stride=1,
padding=1,
name=name + '_conv%d' % (2 * i + 2))
x = x + identity
return x