Create a function dense_net_block that performs the forward pass of a DenseNet dense block on a batch of images stored in an NHWC NumPy tensor input_data (shape (N, H, W, C0)). The block must run num_layers iterations; at each iteration it should (i) apply ReLU to the running feature tensor, (ii) convolve it with the corresponding kernel from kernels (using stride 1, no bias, and symmetric zero-padding of (kh - 1) // 2 rows on the top and bottom and (kw - 1) // 2 columns on the left and right, so that H and W are preserved; kh and kw are odd), and (iii) concatenate the convolution output (whose channel count equals growth_rate) to the running tensor along the channel axis. Every kernel kernels[l] therefore has shape (kh, kw, C0 + l x growth_rate, growth_rate), where (kh, kw) equals kernel_size (default (3, 3)). After the final layer the function must return a tensor of shape (N, H, W, C0 + num_layers x growth_rate). If any kernel's input-channel dimension does not match the current feature-map channels, the function should raise a ValueError.