Implement a PyTorch Conv2d module from tensor operations with configurable channels, kernel, stride, padding, and bias.