openrl.modules.utils package

Submodules

openrl.modules.utils.util module

openrl.modules.utils.util.get_grad_norm(it)

openrl.modules.utils.util.huber_loss(e, d)

openrl.modules.utils.util.mse_loss(e)

openrl.modules.utils.util.update_linear_schedule(optimizer, epoch, total_num_epochs, initial_lr)

Decreases the learning rate linearly

openrl.modules.utils.valuenorm module

class openrl.modules.utils.valuenorm.ValueNorm(input_shape, norm_axes=1, beta=0.99999, per_element_update=False, epsilon=1e-05, device=device(type='cpu'))

Bases: torch.nn.modules.module.Module

Normalize a vector of observations - across the first norm_axes dimensions

denormalize(input_vector)

Transform normalized data back into original distribution

normalize(input_vector)

reset_parameters()

running_mean_var()

update(input_vector)

Module contents