from templates import * from templates_latent import * if __name__ == '__main__': # 256 requires 8x v100s, in our case, on two nodes. # do not run this directly, use `sbatch run_ffhq256.sh` to spawn the srun properly. gpus = [0, 1, 2, 3] nodes = 2 conf = ffhq256_autoenc() train(conf, gpus=gpus, nodes=nodes)