rlightning.models.modules.explore_noise_net

class rlightning.models.modules.explore_noise_net.ExploreNoiseNet(in_dim: int, out_dim: int, hidden_dims: list[int], activation_type: str, noise_logvar_range: list, noise_scheduler_type: str)

Bases: torch.nn.Module

Neural network to generate learnable exploration noise, conditioned on time embeddings and or state embeddings.

forward(noise_feature: MockModule('torch.Tensor'))

post_process(noise_logvar)

input:

torch.Tensor([B, Ta , Da])

output:

torch.Tensor([B, Ta, Da])

set_noise_range(noise_logvar_range: list)

class rlightning.models.modules.explore_noise_net.MLP(dim_list, append_dim=0, append_layers=None, activation_type='tanh', out_activation_type='identity', use_layernorm=False, use_layernorm_final=False, dropout=0, use_drop_final=False, out_bias_init=None, verbose=False)

Bases: torch.nn.Module

forward(x, append=None)