RoboSynChallenge/ACT_sim_drawer_open_place
076
1{2 "type": "act",3 "n_obs_steps": 1,4 "normalization_mapping": {5 "VISUAL": "MEAN_STD",6 "STATE": "MEAN_STD",7 "ACTION": "MEAN_STD"8 },9 "input_features": {10 "observation.state": {11 "type": "STATE",12 "shape": [13 1414 ]15 },16 "observation.qvel": {17 "type": "STATE",18 "shape": [19 1420 ]21 },22 "observation.qf": {23 "type": "STATE",24 "shape": [25 1426 ]27 },28 "observation.images.cam_high": {29 "type": "VISUAL",30 "shape": [31 3,32 480,33 64034 ]35 },36 "observation.images.cam_right_wrist": {37 "type": "VISUAL",38 "shape": [39 3,40 480,41 64042 ]43 },44 "observation.images.cam_left_wrist": {45 "type": "VISUAL",46 "shape": [47 3,48 480,49 64050 ]51 }52 },53 "output_features": {54 "action": {55 "type": "ACTION",56 "shape": [57 1458 ]59 }60 },61 "device": "cuda",62 "use_amp": false,63 "push_to_hub": false,64 "repo_id": null,65 "private": null,66 "tags": null,67 "license": null,68 "chunk_size": 50,69 "n_action_steps": 50,70 "vision_backbone": "resnet18",71 "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",72 "replace_final_stride_with_dilation": false,73 "pre_norm": false,74 "dim_model": 512,75 "n_heads": 8,76 "dim_feedforward": 3200,77 "feedforward_activation": "relu",78 "n_encoder_layers": 4,79 "n_decoder_layers": 1,80 "use_vae": true,81 "latent_dim": 32,82 "n_vae_encoder_layers": 4,83 "temporal_ensemble_coeff": null,84 "dropout": 0.1,85 "kl_weight": 10.0,86 "optimizer_lr": 1e-05,87 "optimizer_weight_decay": 0.0001,88 "optimizer_lr_backbone": 1e-0589}