Atmospheric general circulation models include two main distinct components: the dynamical one solves the Navier-Stokes equations to provide a mathematical representation of atmospheric movements while the physical one includes parameterizations representing small-scale phenomena such as turbulence and convection. However, computational demands of the parameterizations limit the numerical efficiency of the models. Machine learning offers the possibility of developing emulators, as efficient alternatives to traditional parameterizations. We have developed two offline emulators of the physical parameterizations of the IPSL climate model, in an idealized aquaplanet configuration, to reproduce profiles of tendencies of the key variables - zonal wind, meridional wind, temperature, humidity and water tracers - for each atmospheric column. Initial emulators, based on a dense neural network or a convolutional neural network, show good mean performance but struggle with variability. A study of physical processes has revealed that turbulence was at the root of the problem. Knowing how turbulence is parameterized in the model, we show that incorporating physical knowledge through latent variables as predictors into the learning process, leading to a significant improvement of the variability. Future plans involve an online physics emulator, coupled with the atmospheric model to provide a better assessment of the learning process.