def test_CrossEntropyLoss_get_layer_method_with_default_parameters():
x = CrossEntropyLoss()
details = x.get_loss_function()
assert isinstance(details, dict) is True
assert issubclass(details["loss_function"], _CrossEntropyLoss) is True
assert isinstance(details["keyword_arguments"], dict) is True
assert details["keyword_arguments"]["weight"] is None
assert details["keyword_arguments"]["reduction"] == "mean"
assert details["keyword_arguments"]["ignore_index"] == -100
def test_cce_get_layer_method(weight, reduction, ignore_index):
x = CrossEntropyLoss(weight=weight,
reduction=reduction,
ignore_index=ignore_index)
details = x.get_loss_function()
assert isinstance(details, dict) == True
assert issubclass(details["loss_function"], _CrossEntropyLoss) == True
assert isinstance(details["keyword_arguments"], dict) == True
assert torch.all(
torch.eq(details["keyword_arguments"]["weight"],
torch.tensor(weight).float())) == True
assert details["keyword_arguments"]["reduction"] == reduction
assert details["keyword_arguments"]["ignore_index"] == ignore_index
# Create a Sequential model Instance
model = Sequential()
#Build your network
model.add(Conv2D(input_shape=(1, 28, 28), filters=128, kernel_size=3))
model.add(ReLU())
model.add(Conv2D(filters=64, kernel_size=3))
model.add(ReLU())
model.add(Conv2D(filters=32, kernel_size=3))
model.add(ReLU())
model.add(Flatten())
model.add(Dense(n_nodes=10))
model.build()
model.compile(optimizer=SGD(),
loss_function=CrossEntropyLoss(),
metrics=["accuracy"])
print(model.summary())
#Get the MNIST dataset
train_set = torchvision.datasets.MNIST(root='./data',
train=True,
download=True,
transform=transforms.Compose(
[transforms.ToTensor()]))
# Load the dataset from pytorch's Dataloader function
train_loader = torch.utils.data.DataLoader(train_set, batch_size=1000)
#Get the data
mnist_data = next(iter(train_loader))
#Split into train and test set
def test_cce_should_throw_value_error(weight, reduction, ignore_index):
with pytest.raises(ValueError) as ex:
x = CrossEntropyLoss(weight=weight,
reduction=reduction,
ignore_index=ignore_index)
model.add(Conv2D(filters=384, kernel_size=3, stride=1, padding=1))
model.add(ReLU())
model.add(Conv2D(filters=256, kernel_size=3, stride=1, padding=1))
model.add(ReLU())
model.add(MaxPool2D(kernel_size=3, stride=2))
model.add(ReLU())
model.add(Flatten())
model.add(Dense(n_nodes=4096))
model.add(ReLU())
model.add(Dense(n_nodes=4096))
model.add(ReLU())
model.add(Dense(n_nodes=10))
model.add(Softmax())
model.build()
model.compile(optimizer=Adam(), loss_function=CrossEntropyLoss(), metrics=["accuracy"])
print(model.summary())
# Get the training Data
train_set = datasets.MNIST(
root='./data'
,train=True
,download=True
,transform=transforms.Compose([
transforms.CenterCrop(224),
transforms.ToTensor()
])
)
# Load the dataset from pytorch's Dataloader function
train_loader = torch.utils.data.DataLoader(train_set, batch_size=1000)