TODO for next refacto

• factorize SVR and autoencoder
• factorise Square template and Sphere
• Add latent vector as bias (30% speedup)
• remove last th in decoder
• make large .pth tensor with all pointclouds (drop the nasty Chunk_reader)
• make-it multi-gpu
• add netvision results
• rewrite main script object oriented like 3D-CODED
• check that everything works in latest pytorch version
• Add more layer by default and flag for the number of layers
• Add a flag to generate a mesh directly
• Add a python setup install ( that update the submodule, and install the right packages)
• Make sure GPU are used at 100%
• Add f-score in Chamfer + report f-score
• Get rid of shapenet_v2 data and use v1!
• fix path no more sys.path.append
• shapenet 55
• Make minimal dependencies