eCAL

Analytical estimation of the energy cost of the AI lifecycle (J/bit).

eCAL computes the total energy consumed across the full AI model lifecycle — data transmission, preprocessing, training, evaluation, and inference — using closed-form FLOP formulas and hardware power profiles.

Published in IEEE Journal on Selected Areas in Communications (JSAC), 2026.

License

BSD 3-Clause License. See LICENSE.

For contributing/development setup (running tests, linting, type checking), see the project README.