I'm confused

[TheRealJosephTheMan]

so in the video section 54 timestamp 8:02:38 he puts model_1.to(device) next(model_0.parameters()) it didnt worked it makes sense why it didnt work because i didnt create model 1 but i didnt see him create model 1 what do i do? heres colab link: https://colab.research.google.com/drive/1gKl1jetrdNoMU5Brwbz6DFh20BfW6udT?usp=sharing

[tal7aouy]

Based on your google colab code , it seems that you're switching from a custom linear regression model (model_0) to a more straightforward linear model (model_1) using PyTorch's nn.Linear layer. You encountered an issue with model_1 not being created correctly, and there's a reference to using model_1.to(device) which suggests moving the model to a GPU if available.

Addressing the Specific Issue:

1. Initialization of model_1:
   Ensure that model_1 is properly initialized before using it. Your code defines model_1 as an instance of LinearRegressionModelV2:

   class LinearRegressionModelV2(nn.Module):
     def __init__(self):
       super().__init__()
       self.linear_layer = nn.Linear(in_features=1, out_features=1)
   
     def forward(self, x: torch.Tensor) -> torch.Tensor:
         return self.linear_layer(x)

   This model should be created as follows:

   model_1 = LinearRegressionModelV2()

2. Moving Model to Device:
   Before you use model_1.to(device), ensure that device is correctly set:

   device = "cuda" if torch.cuda.is_available() else "cpu"

   Move the model to the device:

   model_1.to(device)

3. Data on the Same Device:
   Ensure that your data (X_train, y_train, X_test, y_test) is also moved to the same device:

   X_train, y_train = X_train.to(device), y_train.to(device)
   X_test, y_test = X_test.to(device), y_test.to(device)

4. Verify Model Parameters:
   After moving the model to the device, verify the device of the model parameters:

   next(model_1.parameters()).device

5. Training Loop:
   Make sure your training loop runs correctly and uses the model on the correct device. Here’s a cleaned-up version of the training loop:

   torch.manual_seed(42)
   epochs = 200
   for epoch in range(epochs):
     model_1.train()
     y_pred = model_1(X_train)
     loss = torch.l1_loss(y_pred, y_train)
     optimiser.zero_grad()
     loss.backward()
     optimiser.step()
     model_1.eval()
     with torch.inference_mode():
       test_pred = model_1(X_test)
       test_loss = torch.l1_loss(test_pred, y_test)
       if epoch % 10 == 0:
         print(f'Epoch: {epoch} | Loss: {loss.item()} | Test loss: {test_loss.item()}')

6. Debugging Output:
   Print the model_1 state_dict and other outputs to ensure everything is working as expected:

   print(model_1.state_dict())

7. Plotting Predictions:
   Ensure that when plotting predictions, you detach tensors from the computation graph and convert them to NumPy arrays:

   with torch.inference_mode():
     y_preds_new = model_1(X_test)
   y_preds_new_np = y_preds_new.detach().cpu().numpy()
   plot_predictions(predictions=y_preds_new_np)

I hope that could help :)

[TheRealJosephTheMan]

thanks