| { |
| "name": "FBcprAuxModel", |
| "device": "cuda", |
| "archi": { |
| "name": "FBcprAuxModelArchiConfig", |
| "z_dim": 256, |
| "norm_z": true, |
| "f": { |
| "name": "ForwardArchi", |
| "hidden_dim": 2048, |
| "model": "residual", |
| "hidden_layers": 6, |
| "embedding_layers": 2, |
| "num_parallel": 2, |
| "ensemble_mode": "batch", |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "privileged_state", |
| "last_action", |
| "history_actor" |
| ] |
| } |
| }, |
| "b": { |
| "name": "BackwardArchi", |
| "hidden_dim": 256, |
| "hidden_layers": 1, |
| "norm": true, |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "privileged_state" |
| ] |
| } |
| }, |
| "actor": { |
| "name": "actor", |
| "model": "residual", |
| "hidden_dim": 2048, |
| "hidden_layers": 6, |
| "embedding_layers": 2, |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "last_action", |
| "history_actor" |
| ] |
| } |
| }, |
| "critic": { |
| "name": "ForwardArchi", |
| "hidden_dim": 2048, |
| "model": "residual", |
| "hidden_layers": 6, |
| "embedding_layers": 2, |
| "num_parallel": 2, |
| "ensemble_mode": "batch", |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "privileged_state", |
| "last_action", |
| "history_actor" |
| ] |
| } |
| }, |
| "discriminator": { |
| "name": "DiscriminatorArchi", |
| "hidden_dim": 1024, |
| "hidden_layers": 3, |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "privileged_state" |
| ] |
| } |
| }, |
| "aux_critic": { |
| "name": "ForwardArchi", |
| "hidden_dim": 1024, |
| "model": "simple", |
| "hidden_layers": 2, |
| "embedding_layers": 2, |
| "num_parallel": 2, |
| "ensemble_mode": "batch", |
| "input_filter": { |
| "name": "DictInputFilterConfig", |
| "key": [ |
| "state", |
| "privileged_state", |
| "last_action", |
| "history_actor" |
| ] |
| } |
| } |
| }, |
| "obs_normalizer": { |
| "name": "ObsNormalizerConfig", |
| "normalizers": { |
| "state": { |
| "name": "BatchNormNormalizerConfig", |
| "momentum": 0.01 |
| }, |
| "privileged_state": { |
| "name": "BatchNormNormalizerConfig", |
| "momentum": 0.01 |
| }, |
| "last_action": { |
| "name": "BatchNormNormalizerConfig", |
| "momentum": 0.01 |
| }, |
| "history_actor": { |
| "name": "BatchNormNormalizerConfig", |
| "momentum": 0.01 |
| } |
| }, |
| "allow_mismatching_keys": true |
| }, |
| "inference_batch_size": 500000, |
| "seq_length": 8, |
| "actor_std": 0.05, |
| "amp": false, |
| "norm_aux_reward": { |
| "name": "RewardNormalizer", |
| "translate": false, |
| "scale": true |
| } |
| } |
