def test_rnn_with_seq_lengths(self, dev=gpu_dev):
bs = 2
seq_length = 3
hidden_size = 2
em_size = 2
x_np = np.array([[[0.1, 0.1], [0.2, 0.2], [0.3, 0.3]],
[[0.3, 0.3], [0.4, 0.4], [0.0,
0.0]]]).astype(np.float32)
y_np = np.array([[0.4, 0.4], [0.5, 0.5]]).astype(np.float32)
seq_lengths_np = np.array([3, 2]).astype(np.int32)
x = tensor.from_numpy(x_np)
x.to_device(dev)
y = tensor.from_numpy(y_np)
y.to_device(dev)
seq_lengths = tensor.from_numpy(seq_lengths_np)
m = LSTMModel3(hidden_size)
m.compile([x, seq_lengths],
is_train=True,
use_graph=False,
sequential=False)
m.train()
for i in range(10):
out = m.forward(x, seq_lengths)
loss = autograd.mse_loss(out, y)
print("train l:", tensor.to_numpy(loss))
m.optimizer(loss)
m.eval()
out = m.forward(x, seq_lengths)
loss = autograd.mse_loss(out, y)
print(" eval l:", tensor.to_numpy(loss))
def test_MeanSquareError(self):
X=np.array([4.3,5.4,3.3,3.6,5.7,6.0]).reshape(3,2).astype(np.float32)
T=np.array([4.4,5.3,3.2,3.7,5.4,6.3]).reshape(3,2).astype(np.float32)
x=tensor.from_numpy(X)
t=tensor.from_numpy(T)
x.to_device(gpu_dev)
t.to_device(gpu_dev)
loss= autograd.mse_loss(x,t)
dx=loss.creator.backward()[0]
loss_np=tensor.to_numpy(loss)[0]
self.assertAlmostEqual(loss_np, 0.0366666, places=4)
self.check_shape(dx.shape(), (3, 2))
def test_MeanSquareError(self):
X=np.array([4.3,5.4,3.3,3.6,5.7,6.0]).reshape(3,2).astype(np.float32)
T=np.array([4.4,5.3,3.2,3.7,5.4,6.3]).reshape(3,2).astype(np.float32)
x=tensor.from_numpy(X)
t=tensor.from_numpy(T)
x.to_device(gpu_dev)
t.to_device(gpu_dev)
loss= autograd.mse_loss(x,t)
dx=loss.creator.backward()[0]
loss_np=tensor.to_numpy(loss)
self.assertAlmostEqual(loss_np, 0.0366666, places=4)
self.check_shape(dx.shape(), (3, 2))
np.random.seed(0)
data = np.random.randn(4, 2).astype(np.float32)
h = np.random.randn(4, 2).astype(np.float32)
label = np.random.randn(4, 2).astype(np.float32)
print('train_data_shape:', data.shape)
print('train_label_shape:', label.shape)
rnn = autograd.RNN(2, 2)
sgd = optimizer.SGD(0.00)
dev = device.get_default_device()
# training process
for i in range(1):
inputs = tensor.Tensor(device=dev, data=data, stores_grad=False)
h0 = tensor.Tensor(device=dev, data=h, stores_grad=False)
targets = tensor.Tensor(device=dev,
data=label,
requires_grad=False,
stores_grad=False)
x = rnn(inputs, h0)
loss = autograd.mse_loss(x, targets)
gradient = autograd.backward(loss)
for p, gp in gradient:
sgd.apply(0, gp, p, '')
if (i % 100 == 0):
print('training loss = ', tensor.to_numpy(loss)[0])
model = sonnx.get_onnx_model(loss, inputs, target)
onnx.save(model, 'linear.onnx')