Neural Network Abstractions Checkpoint

1. The Module class automatically tracks parameters using:

__setattr__ __init__ __call__ __getattr__

2. Xavier initialization uses values in the range:

[-1/sqrt(n), 1/sqrt(n)] [-1, 1] [0, 1] [-0.5, 0.5]

3. Sequential processes modules in:

The order they were added Reverse order Parallel Random order

4. How many parameters does Linear(3, 2, bias=True) have?

5. model.eval() is equivalent to:

model.train(False) model.train(True) model.parameters() model.forward()