-
Notifications
You must be signed in to change notification settings - Fork 16
Description
When I run main.py, I get the following hyperparameters:
Namespace(JFrobint=None, JdiagFrobint=None, JoffdiagFrobint=None, alpha=0.0, atol=1e-05, batch_norm=False, batch_size=1000, bn_lag=0, dataset='data/test_zigzag.npz', dims='64-64-64', divergence_fn='brute_force', dl2int=None, dtl2int=None, embedding_name='gaussians', gpu=1, interp_reg=None, l1int=None, l2int=None, layer_type='concatsquash', leaveout_timepoint=-1, log_freq=10, lr=0.001, max_dim=2, niters=10000, no_display_loss=True, nonlinearity='tanh', num_blocks=1, num_workers=8, rademacher=False, residual=False, rtol=1e-05, save='../trash_results/tmp', save_freq=1000, sl2int=None, solver='dopri5', spectral_norm=False, step_size=None, stochastic=False, test=False, test_atol=None, test_batch_size=1000, test_rtol=None, test_solver=None, time_scale=0.5, top_k_reg=0.0, train_T=True, training_noise=0.1, use_cpu=False, use_density=False, use_growth=False, use_magnitude=False, val_freq=100, vecint=None, viz_batch_size=2000, viz_freq=100, viz_freq_growth=100, weight_decay=1e-05)
Can anyone confirm if these default settings are the correct ones to reproduce the results from the paper? Thanks!