Training deep neural networks requires extensive experimentation to estimate how long it will take, and how much it will cost, for a model to reach convergence. Chronos is a unified framework for pre-training prediction of both the convergence horizon and the resulting end-to-end training time and cost for a target configuration. Unlike approaches that estimate these quantities from observations collected during an executing training job, Chronos uses initialization-time probe signals together with computational, optimization, and hardware descriptors to make predictions before full training begins. Across a diverse spectrum of architectures, from lightweight models such as MobileNetV2 and DeiT-Tiny to deeper and more complex networks, including ViT, ResNet-50, and DenseNet-121, Chronos achieves an average prediction error of 13.7% MAPE for iteration-level execution time and 22.1% MAPE for convergence estimation.