[semester-upkeep] Provide greater detail on nbgrader errors

[jmuchovej]

nbgrader allows us to generate "non-solution" notebooks from "solution" notebooks. More concretely, a Coordinator may write a function like the following:

def train_loop(data: Batches, n_epochs: int = 500):
    # ... some magical preprocessing happens
    for epoch in range(n_epochs):
        # implement the training loop
        ### BEGIN SOLUTION
        for batch in data:
            input, labels = [x.to(device) for x in batch]
            optim.zero_grad()

            output = model(input)
            # some more code
        ### END SOLUTION

and nbgrader will process this cell to produce the following output:

def train_loop(data: Batches, n_epochs: int = 500):
    # ... some magical preprocessing happens
    for epoch in range(n_epochs):
        # implement the training loop
        ### YOUR CODE HERE
        raise NotImplementedError()

Tasks

• (🔑 Required) Retrieve nbgrader's error and dump that to the screen.
• (🍜 Nice to Have) Attempt to correct the error
  • This could actually get gnarly pretty quickly, so this should be saved until the very end.