Model Evaluation positional argument error

[Aditi4AI]

#next (model_0.parameters())
#torch.manual_seed(53)
def eval_model(model:torch.nn.Module,data_loader:torch.utils.data.DataLoader,loss: torch.nn.Module,accuracy):
    Loss,acc=0,0
    model.eval()
    with torch.inference_mode():
        for x,y in data_loader:
            y_pred=model(x)
            Loss+=loss(y_pred,y)
            acc=accuracy (y_true=y,y_pred=y_pred.argmax(dim=1))
        Loss/=len(data_loader)
        acc/=len(data_loader)
    return {"MODEL_NAME": model.__class__.__name__ ,"MODEL_LOSS":Loss.item() ,"model_acc":acc}
model_0_results=eval_model(model=model_0,data_loader=test_dataloader,loss:Loss,accuracy=acc)
model_0_results

It gives the error , what's wrong?

 File "<ipython-input-51-062edb44c824>", line 14
    model_0_results=eval_model(model=model_0,data_loader=test_dataloader,loss:Loss,accuracy=acc)
                                                                             ^
SyntaxError: positional argument follows keyword argument

[mrdbourke]

Hi @Aditi4AI ,

I think Python is getting confused between your variables Loss and loss.

You could try rewriting your function to set loss as loss_fn and Loss as test_loss to escape the confusion.

Also in the line:

model_0_results=eval_model(model=model_0,data_loader=test_dataloader,loss:Loss,accuracy=acc)

You have written loss:Loss, this should be loss=Loss.

But I have updated the whole code to reflect the updates:

#next (model_0.parameters())
#torch.manual_seed(53)
def eval_model(model:torch.nn.Module, data_loader:torch.utils.data.DataLoader, loss_fn: torch.nn.Module, accuracy):
    test_loss, acc = 0, 0
    model.eval()
    with torch.inference_mode():
        for x, y in data_loader:
            y_pred = model(x)
            test_loss += loss_fn(y_pred, y)
            acc = accuracy(y_true=y, y_pred=y_pred.argmax(dim=1))
        test_loss /= len(data_loader)
        acc /= len(data_loader)
    return {"MODEL_NAME": model.__class__.__name__ , "MODEL_LOSS": test_loss.item(), "model_acc": acc}

model_0_results = eval_model(model=model_0, data_loader=test_dataloader, loss_fn=Loss, accuracy=acc)
model_0_results

[Aditi4AI]

Thanks for the help :)
The code executed
I also figured that I have't imported accuracy function from helper function so there was another error also.

[mrdbourke]

You're welcome! Glad you got it fixed! :)