{ "type": "afa", "n_obs_steps": 1, "input_features": { "observation.state": { "type": "STATE", "shape": [ 6 ] }, "observation.images.wrist": { "type": "VISUAL", "shape": [ 3, 256, 256 ] }, "observation.images.front": { "type": "VISUAL", "shape": [ 3, 256, 256 ] }, "observation.vfm_features.wrist": { "type": "STATE", "shape": [ 16, 16, 768 ] }, "observation.vfm_features.wrist_cls": { "type": "STATE", "shape": [ 1, 768 ] }, "observation.vfm_features.front": { "type": "STATE", "shape": [ 16, 16, 768 ] }, "observation.vfm_features.front_cls": { "type": "STATE", "shape": [ 1, 768 ] } }, "output_features": { "action": { "type": "ACTION", "shape": [ 6 ] } }, "device": "cuda", "use_amp": false, "push_to_hub": true, "repo_id": "Tsagkas/afa_v0", "private": null, "tags": null, "license": null, "pretrained_path": null, "use_afa": true, "feature_map_shape": [ 16, 16, 768 ], "state_dim": 6, "action_dim": 6, "chunk_size": 100, "n_action_steps": 1, "temporal_ensemble_coeff": 0.01, "normalization_mapping": { "VISUAL": "IDENTITY", "STATE": "MEAN_STD", "ACTION": "MEAN_STD" }, "dim_model": 768, "n_heads": 8, "mlp_hidden_dim": 512, "mlp_num_layers": 5, "dropout": 0.1, "optimizer_lr": 0.0001, "optimizer_weight_decay": 0.0001 }