In recent years, the use of Artificial Intelligence (AI) has experienced exponential growth. However, this development has also raised a new concern: the high energy consumption associated with the life cycle of its models and their environmental impact. Deep learning, and in particular convolutional networks, are among the models that consume the most computational resources during their training.

To address this challenge, we present GREENN (Granular evaluation of Energy Efficiency in Neural Networks), a tool designed to help Machine Learning (ML) practitioners understand the energy behavior of their neural networks and choose the most appropriate architecture for their specific problem, thus achieving a balance between performance and energy consumption. To achieve this, GREENN measures and analyzes energy consumption during training at different levels of granularity: (i) taking into account the overall training process, (ii) breaking down the results for each epoch, or (iii) for each of the layers of the neural network. In addition to tracking energy usage per hardware component and associated carbon emissions, GREENN provides model performance metrics such as accuracy and F1-score, enabling a comprehensive evaluation that considers both computational efficiency and predictive capability.