Feedback about Warm-up: numpy - on the Polynomial (Mis)Fit Example in the Tensors Tutorial

Hi there,

I was looking at the tutorial on fitting a polynomial with NumPy and PyTorch:
https://docs.pytorch.org/tutorials/beginner/examples_tensor/polynomial_numpy.html

I noticed that the resulting third-degree polynomial is not a very good fit for the sin(x) function it's supposed to approximate:
f(x) = 0.0003655185795843861 + 0.8513230793867113 x + -6.305803723372667e-05 x² + -0.09255975933441395 x³

You can see a graph of both functions here—they look pretty different:
https://www.geogebra.org/calculator/qyykjrjf

My concern is that this might be a confusing example for beginners. It could give them the wrong impression of what a good result looks like when using PyTorch.

Cheers,
Carlos Rogério Camargo

cc @albanD @jbschlosser

Start with the polynomial NumPy tutorial at docs.pytorch.org/tutorials/beginner/examples_tensor/polynomial_numpy.html and reproduce the reported comparison with sin(x). Check whether the example or its explanation needs correction, and consider the work complete when the tutorial gives beginners an accurate, clearly explained interpretation of the fit.