vagan.hidden_dim = FLAGS.hidden_dim vagan.dataname_a = FLAGS.dataname_a vagan.dataname_b = FLAGS.dataname_b vagan.test_vae = FLAGS.test_vae vagan.test_gan = FLAGS.test_gan vagan.tb = FLAGS.tb vagan.lambda_2 = FLAGS.lambda_2 vagan.vae_ablity = FLAGS.vae_ablity vagan.logfolder = FLAGS.logfolder os.environ["CUDA_VISIBLE_DEVICES"] = str(FLAGS.gpu) train_model = vagan.convMESH() with tf.Session(config=train_model.config) as train_model.sess: train_model.train_pre() train_VAE(train_model) train_metric(train_model) ''' #_model.train_VAE() _model.test_vae(1) # _model.test_vae_itlp(1) # _model.test_metric(1) # _model.recons_error_a() # _model.recons_error_b() '''
vcgan.hidden_dim = FLAGS.hidden_dim vcgan.dataname_a = FLAGS.dataname_a #vcgan.dataname_b = FLAGS.dataname_b #vcgan.test_vae = FLAGS.test_vae #vcgan.test_gan = FLAGS.test_gan vcgan.tb = FLAGS.tb vcgan.lambda_2 = FLAGS.lambda_2 vcgan.vae_ablity = FLAGS.vae_ablity vcgan.logfolder = FLAGS.logfolder vcgan.featurefile_a = './Features0412.mat' # for faster initialization os.environ["CUDA_VISIBLE_DEVICES"] = str(FLAGS.gpu) #os.environ["CUDA_VISIBLE_DEVICES"] = "9" train_model = vcgan.convMESH() feature_a, neighbour1_a, degree1_a, logrmin_a, logrmax_a, smin_a, smax_a, modelnum_a, \ pointnum1_a, maxdegree1_a, L1_a, cotw1_a = utils.load_data(vcgan.featurefile_a, vcgan.resultmin, vcgan.resultmax, useS=vcgan.useS) with tf.Session(config=train_model.config) as train_model.sess: # train_model.train_pre() train_model.load_test('VAE_Features0412_16_128/vae_a.model-200') # train_vae.train_VAE(train_model) # train_metric(train_model) test_utils.recons_error(train_model,feature_a) ''' #_model.train_VAE() _model.test_vae(1) # _model.test_vae_itlp(1) # _model.test_metric(1)
parser.add_argument('-l', '--layers', default=1, type=int) # parser.add_argument('-m', '--maxepoch', default=2000, type = str) parser.add_argument('--modelfile', default='convmesh-model-2000', type=str) args = parser.parse_args() hidden_dim = args.hiddendim featurefile = args.featurefile neighbourfile = args.neighbourfile neighbourvariable = args.neighbourvariable # distancefile = args.distancefile # distancevariable = args.distancevariable lambda1 = args.l1 lambda2 = args.l2 lr = args.lr finaldim = args.finaldim layers = args.layers modelfile = args.modelfile # maxepoch = args.maxepoch feature, logrmin, logrmax, smin, smax, pointnum = load_data(featurefile) neighbour, degrees, maxdegree = load_neighbour(neighbourfile, neighbourvariable, pointnum) # geodesic_weight = load_geodesic_weight(distancefile, distancevariable, pointnum) model = model.convMESH(pointnum, neighbour, degrees, maxdegree, hidden_dim, finaldim, layers, lambda1, lambda2, lr) model.individual_dimension(modelfile, feature, logrmin, logrmax, smin, smax)
model.hidden_dim = FLAGS.hidden_dim model.dataname_a = FLAGS.dataname_a model.dataname_b = FLAGS.dataname_b model.test_vae = FLAGS.test_vae model.test_gan = FLAGS.test_gan model.tb = FLAGS.tb model.lambda_2 = FLAGS.lambda_2 model.vae_ablity = FLAGS.vae_ablity model.logfolder = FLAGS.logfolder os.environ["CUDA_VISIBLE_DEVICES"] = str(FLAGS.gpu) print(model.dataname_a) train_model = model.convMESH() with tf.Session(config=train_model.config) as train_model.sess: #''' train_model.train_pre() train_VAE(train_model) train_metric(train_model) train_GAN(train_model) ''' train_model.train_pre() test_vae(train_model,30000) test_metric(train_model,22000) test_gan(train_model,30000) '''