Section 09 Lecture 207 seems to be missing instruction from earlier, or the copied in code doesn't work

[AxisMeetsWorld]

Hi @mrdbourke,
So, I ran the code as I think you specified in the previous lecture 206, and caught an error or two, but I am wondering if this
import is really working as expected: from going_modular.going_modular.engine import train_step, test_step

I got to about the 1:20 point in your video on lecture 207, and almost nothing from that point works. You didn't really specify what library "set_seeds()" is from, and everytime I try to run that I am getting an error: NameError: name 'set_seeds' is not defined.

I commented that out to see if perhaps I could get this to run on its own:
results = train(model=model, train_dataloader=train_dataloader, test_dataloader=test_dataloader, optimizer=optimizer, loss_fn=loss_fn, epochs=5, device=device)
That didn't work either, as I ended up getting a

• RuntimeError: Boolean value of Tensor with more than one value is ambiguous * error. It looked like it was from the nn.reduction.py library. I searched the code in the previous cell where we were reworking the .train() function but importing train_step, and test_step but the only thing that I could see going wrong was where we tried to import in. Is there some other way we need to be pulling this in? It seemed the suggestions that cropped up seemed to support using "/" but what puzzles me is that period separated format seems to work for you. Please see the screenshot below, and help me determine what could possibly be going wrong to get these errors.

Here is the error 👎

[AxisMeetsWorld]

Say, I found the definition of the set_seeds() from the call you had earlier. I didn't see that somehow in the videos so apologies if I got that wrong 😬, but I am still lost on the rest. What could be causing that error?

[AxisMeetsWorld]

@mrdbourke I looked more at this today, and even went through line by line with the book version and I am still getting this error 😞 . RuntimeError: Boolean value of Tensor with more than one value is ambiguous However, when I did look at the error more, it looks like there might be some sort of issue where the variables reduce and size_average are expected to return some sort of true/false, and perhaps the fact that they are set to None could be throwing them off?
Here is the error code, and I am wondering if there is some update I missed or failed to implement on this chapter? Also, if there is anyone else I should tag on these, please let me know. I don't want to bombard just you if there are others that can help. Please help with this one because I have dug up as much as I can:

[<ipython-input-13-88af18dc4fb2>](https://localhost:8080/#) in train(model, train_dataloader, test_dataloader, optimizer, loss_fn, epochs, device)
     54     # Loop through training and testing steps for a number of epochs
     55     for epoch in tqdm(range(epochs)):
---> 56         train_loss, train_acc = train_step(model=model,
     57                                           dataloader=train_dataloader,
     58                                           loss_fn=loss_fn,

[/content/going_modular/going_modular/engine.py](https://localhost:8080/#) in train_step(model, dataloader, loss_fn, optimizer, device)
     46 
     47         # 2. Calculate  and accumulate loss
---> 48         loss = loss_fn(y_pred, y)
     49         train_loss += loss.item()
     50 

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/loss.py](https://localhost:8080/#) in __init__(self, weight, size_average, ignore_index, reduce, reduction, label_smoothing)
   1167     def __init__(self, weight: Optional[Tensor] = None, size_average=None, ignore_index: int = -100,
   1168                  reduce=None, reduction: str = 'mean', label_smoothing: float = 0.0) -> None:
-> 1169         super().__init__(weight, size_average, reduce, reduction)
   1170         self.ignore_index = ignore_index
   1171         self.label_smoothing = label_smoothing

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/loss.py](https://localhost:8080/#) in __init__(self, weight, size_average, reduce, reduction)
     28 class _WeightedLoss(_Loss):
     29     def __init__(self, weight: Optional[Tensor] = None, size_average=None, reduce=None, reduction: str = 'mean') -> None:
---> 30         super().__init__(size_average, reduce, reduction)
     31         self.register_buffer('weight', weight)
     32         self.weight: Optional[Tensor]

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/loss.py](https://localhost:8080/#) in __init__(self, size_average, reduce, reduction)
     21         super().__init__()
     22         if size_average is not None or reduce is not None:
---> 23             self.reduction: str = _Reduction.legacy_get_string(size_average, reduce)
     24         else:
     25             self.reduction = reduction

[/usr/local/lib/python3.10/dist-packages/torch/nn/_reduction.py](https://localhost:8080/#) in legacy_get_string(size_average, reduce, emit_warning)
     33         reduce = True
     34 
---> 35     if size_average and reduce:
     36         ret = 'mean'
     37     elif reduce:

[AxisMeetsWorld]

Ok, I am a little 😳 shamed....I should have looked here. I missed that on the call loss_fn = nn.CrossEntropyLoss() I hadn't placed the parenthesis down. Unfortunately, the error message wasn't too helpful here. Putting those in solve the problem, and the rest of the code works now.