In the last years, neural networks have been used to learn physical simulations in a wide range of contexts. The present work tackles the training of neural networks for large deformation plasticity. There are two sources of nonlinearity: geometric (large deformation) and material (plasticity). Traditional numerical methods for plastic simulations (such as the Finite Element method) are computationally expensive. NNs architectures have been proposed in plasticity in some simple cases, in the small deformations framework. The main contributions are i) the application of NN for plasticity in large deformations, ii) a review and comparison of the existing methods, iii) the use of a Temporal Convolutional Network that trains faster than the existing methods, iv) an open source benchmark to the scientific community. Altogether, these contributions enable real-time simulation of plastic behaviors.