Backprop Through a Two-Layer Graph

A tiny network computes its output through this chain of operations:

$z = w_1 x + b_1, \qquad a = \sigma(z), \qquad \hat{y} = w_2 a, \qquad L = \tfrac{1}{2}(\hat{y} - y)^2$

where $\sigma(z) = \dfrac{1}{1 + e^{-z}}$ is the sigmoid.

The current values are

symbol	$x$	$w_1$	$b_1$	$w_2$	$y$
value	$2$	$1.5$	$-3$	$4$	$3$

Run the forward pass, then backpropagate. What is $\dfrac{\partial L}{\partial w_1}$ at these values?