Upload config.json with huggingface_hub
Browse files- config.json +168 -0
config.json
ADDED
|
@@ -0,0 +1,168 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "hybrid",
|
| 3 |
+
"n_obs_steps": 1,
|
| 4 |
+
"normalization_mapping": {},
|
| 5 |
+
"input_features": {
|
| 6 |
+
"observation.images.top": {
|
| 7 |
+
"type": "VISUAL",
|
| 8 |
+
"shape": [
|
| 9 |
+
3,
|
| 10 |
+
480,
|
| 11 |
+
640
|
| 12 |
+
]
|
| 13 |
+
},
|
| 14 |
+
"observation.state": {
|
| 15 |
+
"type": "STATE",
|
| 16 |
+
"shape": [
|
| 17 |
+
14
|
| 18 |
+
]
|
| 19 |
+
}
|
| 20 |
+
},
|
| 21 |
+
"output_features": {
|
| 22 |
+
"action": {
|
| 23 |
+
"type": "ACTION",
|
| 24 |
+
"shape": [
|
| 25 |
+
14
|
| 26 |
+
]
|
| 27 |
+
}
|
| 28 |
+
},
|
| 29 |
+
"device": "cuda",
|
| 30 |
+
"use_amp": false,
|
| 31 |
+
"act_config": {
|
| 32 |
+
"type": "act",
|
| 33 |
+
"n_obs_steps": 2,
|
| 34 |
+
"normalization_mapping": {
|
| 35 |
+
"VISUAL": "MEAN_STD",
|
| 36 |
+
"STATE": "MEAN_STD",
|
| 37 |
+
"ACTION": "MEAN_STD"
|
| 38 |
+
},
|
| 39 |
+
"input_features": {
|
| 40 |
+
"observation.state": {
|
| 41 |
+
"type": "STATE",
|
| 42 |
+
"shape": [
|
| 43 |
+
14
|
| 44 |
+
]
|
| 45 |
+
},
|
| 46 |
+
"observation.images.top": {
|
| 47 |
+
"type": "VISUAL",
|
| 48 |
+
"shape": [
|
| 49 |
+
3,
|
| 50 |
+
480,
|
| 51 |
+
640
|
| 52 |
+
]
|
| 53 |
+
}
|
| 54 |
+
},
|
| 55 |
+
"output_features": {
|
| 56 |
+
"action": {
|
| 57 |
+
"type": "ACTION",
|
| 58 |
+
"shape": [
|
| 59 |
+
14
|
| 60 |
+
]
|
| 61 |
+
}
|
| 62 |
+
},
|
| 63 |
+
"device": "cuda",
|
| 64 |
+
"use_amp": false,
|
| 65 |
+
"chunk_size": 400,
|
| 66 |
+
"n_action_steps": 400,
|
| 67 |
+
"vision_backbone": "resnet18",
|
| 68 |
+
"pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
|
| 69 |
+
"replace_final_stride_with_dilation": false,
|
| 70 |
+
"pre_norm": false,
|
| 71 |
+
"dim_model": 512,
|
| 72 |
+
"n_heads": 8,
|
| 73 |
+
"dim_feedforward": 3200,
|
| 74 |
+
"feedforward_activation": "relu",
|
| 75 |
+
"n_encoder_layers": 4,
|
| 76 |
+
"n_decoder_layers": 1,
|
| 77 |
+
"use_vae": true,
|
| 78 |
+
"latent_dim": 32,
|
| 79 |
+
"n_vae_encoder_layers": 4,
|
| 80 |
+
"temporal_ensemble_coeff": null,
|
| 81 |
+
"dropout": 0.1,
|
| 82 |
+
"kl_weight": 10.0,
|
| 83 |
+
"smoothness_weight": 0.0,
|
| 84 |
+
"smoothness_type": "velocity",
|
| 85 |
+
"optimizer_lr": 1e-05,
|
| 86 |
+
"optimizer_weight_decay": 0.0001,
|
| 87 |
+
"optimizer_lr_backbone": 1e-05
|
| 88 |
+
},
|
| 89 |
+
"diffusion_config": {
|
| 90 |
+
"type": "diffusion",
|
| 91 |
+
"n_obs_steps": 2,
|
| 92 |
+
"normalization_mapping": {
|
| 93 |
+
"VISUAL": "MEAN_STD",
|
| 94 |
+
"STATE": "MEAN_STD",
|
| 95 |
+
"ACTION": "MEAN_STD"
|
| 96 |
+
},
|
| 97 |
+
"input_features": {
|
| 98 |
+
"observation.state": {
|
| 99 |
+
"type": "STATE",
|
| 100 |
+
"shape": [
|
| 101 |
+
14
|
| 102 |
+
]
|
| 103 |
+
},
|
| 104 |
+
"observation.images.top": {
|
| 105 |
+
"type": "VISUAL",
|
| 106 |
+
"shape": [
|
| 107 |
+
3,
|
| 108 |
+
480,
|
| 109 |
+
640
|
| 110 |
+
]
|
| 111 |
+
}
|
| 112 |
+
},
|
| 113 |
+
"output_features": {
|
| 114 |
+
"action": {
|
| 115 |
+
"type": "ACTION",
|
| 116 |
+
"shape": [
|
| 117 |
+
14
|
| 118 |
+
]
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"device": "cuda",
|
| 122 |
+
"use_amp": false,
|
| 123 |
+
"horizon": 16,
|
| 124 |
+
"n_action_steps": 1,
|
| 125 |
+
"drop_n_last_frames": 7,
|
| 126 |
+
"vision_backbone": "resnet18",
|
| 127 |
+
"crop_shape": null,
|
| 128 |
+
"crop_is_random": true,
|
| 129 |
+
"pretrained_backbone_weights": null,
|
| 130 |
+
"use_group_norm": true,
|
| 131 |
+
"spatial_softmax_num_keypoints": 32,
|
| 132 |
+
"use_separate_rgb_encoder_per_camera": false,
|
| 133 |
+
"down_dims": [
|
| 134 |
+
128,
|
| 135 |
+
256,
|
| 136 |
+
512
|
| 137 |
+
],
|
| 138 |
+
"kernel_size": 5,
|
| 139 |
+
"n_groups": 8,
|
| 140 |
+
"diffusion_step_embed_dim": 128,
|
| 141 |
+
"use_film_scale_modulation": true,
|
| 142 |
+
"noise_scheduler_type": "DDIM",
|
| 143 |
+
"num_train_timesteps": 100,
|
| 144 |
+
"beta_schedule": "squaredcos_cap_v2",
|
| 145 |
+
"beta_start": 0.0001,
|
| 146 |
+
"beta_end": 0.02,
|
| 147 |
+
"prediction_type": "epsilon",
|
| 148 |
+
"clip_sample": true,
|
| 149 |
+
"clip_sample_range": 1.0,
|
| 150 |
+
"num_inference_steps": 80,
|
| 151 |
+
"do_mask_loss_for_padding": false,
|
| 152 |
+
"optimizer_lr": 0.0001,
|
| 153 |
+
"optimizer_betas": [
|
| 154 |
+
0.95,
|
| 155 |
+
0.999
|
| 156 |
+
],
|
| 157 |
+
"optimizer_eps": 1e-08,
|
| 158 |
+
"optimizer_weight_decay": 1e-06,
|
| 159 |
+
"scheduler_name": "cosine",
|
| 160 |
+
"scheduler_warmup_steps": 500
|
| 161 |
+
},
|
| 162 |
+
"act_loss_weight": 1.0,
|
| 163 |
+
"diffusion_loss_weight": 1.0,
|
| 164 |
+
"num_keyframes_to_refine": 15,
|
| 165 |
+
"refinement_window_past": 4,
|
| 166 |
+
"refinement_window_future": 7,
|
| 167 |
+
"action_blend_alpha": 0.4
|
| 168 |
+
}
|