Towards geological inference with process-based and deep generative modeling, part 1: training on fluvial deposits