 The proposed MT-Cybe model is based on the idea of compressing the temporal dimension of time series data. This compression is achieved through the use of partial convolutions, which are similar to the operations used in image processing. The model is then evaluated against traditional time series models, showing improved accuracy and efficiency. Additionally, the model is shown to be applicable to multiple datasets, including electricity production, road traffic, and astronomy. This article was authored by Dennis Oman, Olga Taran, and Slava Voloshinovsky.