def apply(self, *variables_list):
ctx = Autograd.Context()
input_tensors = [v.data for v in variables_list]
forward_tensor = self.forward(ctx, *input_tensors)
output_variable = Variable(forward_tensor)
output_variable.backward_function = lambda x: self.backward(ctx, x)
output_variable.backward_variables = [v for v in variables_list]
return output_variable
def __init__(self, num_input, num_output):
super().__init__()
self.num_input = num_input
self.num_output = num_output
self.weights = Variable(
np.random.normal(0, 0.05, (self.num_input, self.num_output)))
self.biases = Variable(np.zeros(self.num_output, dtype=np.float32))
self.register_variables(self.weights, self.biases)
def apply(cls, *variables_list):
new_context = Context()
input_tensors = [v.data for v in variables_list]
forward_tensor = cls.forward(new_context, *input_tensors)
output_variable = Variable(forward_tensor)
output_variable.backward_function = lambda x: cls.backward(
new_context, x)
output_variable.backward_variables = [v for v in variables_list]
return output_variable
def __init__(self,
input_channels,
output_channels,
kernel_size,
stride=1,
padding=0):
super().__init__()
if isinstance(kernel_size, int):
kernel_size = (kernel_size, kernel_size)
if isinstance(stride, int):
stride = (stride, stride)
self.input_channels = input_channels
self.output_channels = output_channels
self.kernel_size = kernel_size
self.padding = padding
self.stride = stride
self.weights = Variable(
np.random.normal(0, 0.05,
(self.input_channels, self.output_channels,
self.kernel_size[0], self.kernel_size[1])))
self.biases = variable.zeros(output_channels)
def test_model_acc():
correct = 0
for test_batch_in, test_batch_out in test_batches:
test_output = my_model(Variable(test_batch_in)).data
correct += np.sum(np.argmax(test_output, axis=1) == np.argmax(test_batch_out, axis=1))
return correct / len(test_dataset)
for test_batch_in, test_batch_out in test_batches:
test_output = my_model(Variable(test_batch_in)).data
correct += np.sum(np.argmax(test_output, axis=1) == np.argmax(test_batch_out, axis=1))
my_acc = correct / len(test_dataset)
return my_acc
finished = False
for it in range(iterations):
if finished:
break
train_batches.shuffle()
for i_b, (batch_in, batch_out) in enumerate(train_batches):
model_input = Variable(batch_in)
good_output = Variable(batch_out)
model_output = my_model(model_input)
err = loss(good_output, model_output)
optimizer.zero_grad()
err.backward()
optimizer.step()
acc = test_model_acc()
print("model accuracy: {}".format(acc))
if acc > 0.97:
finished = True