Counting parameters in a 2D convolution layer (with bias)

In PyTorch you create the layer `nn.Conv2d(in_channels=16, out_channels=32, kernel_size=3, stride=1, padding=1, bias=True)`. Assuming a standard (non-grouped, `groups=1`) convolution, how many learnable parameters does this single layer have? Count both weights and biases. Give your answer as a single integer (number of parameters).