Video "PyTorch for Deep Learning & Machine Learning – Full Course" Coding error in model training and testing

[Rohi2610]

I was training my model and I ran into an error I'm unable to resolve. The error is showing in the training forward pass but I think the actual problem is in the LinearRegressionModelV2 class, more accurately in the forward function of the class. However, I'm unable to pinpoint the error. Here's a picture of the error

Here's my code for the LinearRegressionModelV2 class

# Building a linear model
class LinearRegressionModelV2(nn.Module):
  def __init__(self):
    super().__init__()
    # Using nn.linear() since we are building a linear model
    self.linear_layer = nn.Linear(in_features=1, 
                                  out_features=1)
    
  def forward(self, x: torch.Tensor) -> torch.Tensor:
    return self.linear_layer(x)
torch.manual_seed(42)
model_1 = LinearRegressionModelV2()
model_1, model_1.state_dict()

Here's the training and testing loop

# Loss function
loss_fn1 = nn.L1Loss()

# Optimizer
optimizer = torch.optim.SGD(params = model_1.parameters(),
                            lr = 0.01)

# training and testing Loop
torch.manual_seed(42)

epoch = 200

# Put data on target device(device agnostic code)
X_train = X_train.to(device)
Y_train = Y_train.to(device)
X_test = X_test.to(device)
Y_test = Y_test.to(device)

for epochs in range(epoch):
  model_1.train()

  # Forward pass
  y_pred = model_1(X_train)

  # Calculate loss
  loss = loss_fn1(y_pred, Y_train)

  # Optimizer zero grad
  optimizer.zero_grad()

  # Perform Backpropagation
  loss.backward()

  # Optimizer step
  optimizer.step()


  # Testing
  model_1.eval()
  with torch.inference_mode():
    test_pred = model_1(X_test)

    test_loss = loss_fn1(test_pred, Y_test)

  # Printing
  if epoch % 10 == 0:
    print(f"Epoch: {epochs} | Loss: {loss} | Test Loss: {test_loss}")

[LuluW8071]

The snippet of the code you provided has no errors works perfectly fine. I tried to replicate your error and closest i found was on defining weight and bias. There might be error(mistype) on that section.

weights = 0.7
bias = 0.3
start, end, step = 0, 1, 0.02

X = torch.arange(start, end, step).unsqueeze(dim=1)
y = weights * X + bias

[Rohi2610]

Thank you so much. Turns out I did forget to unsqueeze X. It's working well.