Context. Measuring how the physical properties of galaxies change across cosmic times is essential to understand galaxy formation and evolution. With the advent of numerous ground-based and space-borne instruments launched over the past few decades we now have exquisite multi-wavelength observations of galaxies from the FUV to the radio domain. To tap into this mine of data and obtain new insight into the formation and evolution of galaxies, it is essential that we are able to extract information from their SED. Aims. We present a completely new implementation of CIGALE. Written in python, its main aims are to easily and efficiently model the FUV to radio spectrum of galaxies and estimate their physical properties such as star formation rate, attenuation, dust luminosity, stellar mass, and many other physical quantities. Methods. To compute the spectral models, CIGALE builds composite stellar populations from simple stellar populations combined with highly flexible star formation histories, calculates the emission from gas ionised by massive stars, and attenuates both the stars and the ionised gas with a highly flexible attenuation curve. Based on an energy balance principle, the absorbed energy is then re-emitted by the dust in the mid- and far-infrared domains while thermal and non-thermal components are also included, extending the spectrum far into the radio range. A large grid of models is then fitted to the data and the physical properties are estimated through the analysis of the likelihood distribution. Results. CIGALE is a versatile and easy-to-use tool that makes full use of the architecture of multi-core computers, building grids of millions of models and analysing samples of thousands of galaxies, both at high speed. Beyond fitting the SEDs of galaxies and parameter estimations, it can also be used as a model-generation tool or serve as a library to build new applications.

Dust attenuation shapes the spectral energy distribution of galaxies. It is particularly true for dusty galaxies in which stars experience a heavy attenuation. The combination of UV-to-IR photometry with the spectroscopic measurement of the H$\alpha$ recombination line helps to quantify dust attenuation of the whole stellar population and its wavelength dependence. We selected an IR complete sample of galaxies in the COSMOS 3D-HST CANDELS field detected with the Herschel satellite with a signal to noise ratio larger than five. Optical to NIR photometry is available as well as NIR spectroscopy for each source. We reduced the sample to the redshift range $0.6 < z < 1.6$ to include the H$\alpha$ line in the G141 grism spectra. We have used a new version of the CIGALE code to fit simultaneously the continuum and H$\alpha$ line emission of the 34 selected galaxies. Using flexible attenuation laws with free parameters, we are able to measure the shape of the attenuation curve for each galaxy as well as the amount of attenuation of each stellar population, the former being in general steeper than the starburst law in the UV-optical with a large variation of the slope among galaxies. The attenuation of young stars or nebular continuum is found on average about twice the attenuation affecting older stars, again with a large variation. Our model with power-laws, based on a modification of the Charlot and Fall recipe, gives results in better agreement with the radiative transfer models than the global modification of the slope of the Calzetti law.