Hi
i’m still new to Pytorch and currently using my own dataset.
For some reason, i have huge problems with getting the dimensions right for my network.
The data is being laoded from a .csv file in the beginning.
This is my code so far:
X_train, X_test, y_train, y_test = train_test_split(df, y, test_size=0.2)
train_set = TensorDataset(torch.from_numpy(X_train.values),
torch.from_numpy(y_train.values))
test_set = TensorDataset(torch.from_numpy(X_test.values),
torch.from_numpy(y_test.values))
train_loader = data_utils.DataLoader(train_set,
batch_size=16,
shuffle=True)
test_loader = data_utils.DataLoader(test_set,
batch_size=16,
shuffle=False)
input_dim = 11
num_classes = 25
class Net(torch.nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv1d(input_dim, 64, kernel_size=2)
self.conv2 = nn.Conv1d(64, 64, kernel_size=2)
self.pool = nn.MaxPool2d(kernel_size=2)
self.lin1 = torch.nn.Linear(64, 32)
self.lin2 = torch.nn.Linear(32, int(num_classes))
def forward(self, data):
x = data
print(x.shape)
print(self.conv1)
#x = self.norm(x)
x = F.relu(self.conv1(x))
x = F.dropout(x, training=self.training)
x = F.relu(self.conv2(x))
x = self.pool(x)
x = F.relu(self.lin1(x))
x = F.relu(self.lin2(x))
return F.log_softmax(x, dim=1)
model = Net()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
def train():
model.train()
total_loss = 0
correct = 0
for data, label in train_loader:
optimizer.zero_grad()
print(data.shape, label.shape)
loss = F.nll_loss(model(data),label.long())
loss.backward()
total_loss += loss.item()
with torch.no_grad():
pred = model(data).max(dim=1)[1]
correct += pred.eq(label.long()).sum().item()
optimizer.step()
return total_loss / len(train), correct / len(train)
def test(loader):
model.eval()
correct = 0
for data, label in test_loader:
with torch.no_grad():
pred = model(data).max(dim=1)[1]
correct += pred.eq(label.long()).sum().item()
return correct / len(loader)
epochs = 100
for epoch in range(1, epochs + 1):
loss, train_acc = train()
test_acc = test(test_loader)
print('Epoch {:03d}, Train Loss: {:.4f}, Train Accuracy: {:.4f}, Test Accuracy: {:.4f}'.format(
epoch, loss, train_acc, test_acc))
The data shape results for x and y are
torch.Size([16, 11]) torch.Size([16])
And i’m getting the following error:
File "C:\Users\Network.py", line 75, in forward
x = F.relu(self.conv1(x))
File "C:\Users\AppData\Local\conda\conda\envs\Analysis\lib\site-packages\torch\nn\modules\module.py", line 541, in __call__
result = self.forward(*input, **kwargs)
File "C:\Users\AppData\Local\conda\conda\envs\Analysis\lib\site-packages\torch\nn\modules\conv.py", line 202, in forward
self.padding, self.dilation, self.groups)
RuntimeError: Expected 3-dimensional input for 3-dimensional weight 64 11 2, but got 2-dimensional input of size [16, 11] instead
Also: here is an example, of how a single line of my data looks:
data = [1,2,3,4,5,6,7,8,9,10,11]
y = [1]
I’m really stumped, as to what my problem is.
Any help would be really appreciated