feat: Add pipeline config to load from hf directly.
#1
by
Kiiimm - opened
post_training_robotwin/robotwin2_0.config.json
ADDED
|
@@ -0,0 +1,190 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"load_depth": true,
|
| 10 |
+
"cam_names": [
|
| 11 |
+
"front_camera",
|
| 12 |
+
"left_camera",
|
| 13 |
+
"right_camera",
|
| 14 |
+
"head_camera"
|
| 15 |
+
],
|
| 16 |
+
"valid_action_step": null,
|
| 17 |
+
"transforms": [
|
| 18 |
+
{
|
| 19 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddItems",
|
| 20 |
+
"T_base2world": [
|
| 21 |
+
[
|
| 22 |
+
0,
|
| 23 |
+
-1,
|
| 24 |
+
0,
|
| 25 |
+
0
|
| 26 |
+
],
|
| 27 |
+
[
|
| 28 |
+
1,
|
| 29 |
+
0,
|
| 30 |
+
0,
|
| 31 |
+
-0.65
|
| 32 |
+
],
|
| 33 |
+
[
|
| 34 |
+
0,
|
| 35 |
+
0,
|
| 36 |
+
1,
|
| 37 |
+
0
|
| 38 |
+
],
|
| 39 |
+
[
|
| 40 |
+
0,
|
| 41 |
+
0,
|
| 42 |
+
0,
|
| 43 |
+
1
|
| 44 |
+
]
|
| 45 |
+
],
|
| 46 |
+
"joint_mask": [
|
| 47 |
+
true,
|
| 48 |
+
true,
|
| 49 |
+
true,
|
| 50 |
+
true,
|
| 51 |
+
true,
|
| 52 |
+
true,
|
| 53 |
+
false,
|
| 54 |
+
true,
|
| 55 |
+
true,
|
| 56 |
+
true,
|
| 57 |
+
true,
|
| 58 |
+
true,
|
| 59 |
+
true,
|
| 60 |
+
false
|
| 61 |
+
]
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:SimpleStateSampling",
|
| 65 |
+
"hist_steps": 1,
|
| 66 |
+
"pred_steps": 64
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:Resize",
|
| 70 |
+
"dst_wh": [
|
| 71 |
+
320,
|
| 72 |
+
256
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ImageChannelFlip",
|
| 77 |
+
"output_channel": [
|
| 78 |
+
2,
|
| 79 |
+
1,
|
| 80 |
+
0
|
| 81 |
+
]
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ToTensor"
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:MoveEgoToCam"
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:GetProjectionMat",
|
| 91 |
+
"target_coordinate": "ego"
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddScaleShift",
|
| 95 |
+
"scale_shift": [
|
| 96 |
+
[
|
| 97 |
+
1.12735104,
|
| 98 |
+
-0.11648428
|
| 99 |
+
],
|
| 100 |
+
[
|
| 101 |
+
1.45046443,
|
| 102 |
+
1.35436516
|
| 103 |
+
],
|
| 104 |
+
[
|
| 105 |
+
1.5324732,
|
| 106 |
+
1.45750941
|
| 107 |
+
],
|
| 108 |
+
[
|
| 109 |
+
1.80842297,
|
| 110 |
+
-0.01855904
|
| 111 |
+
],
|
| 112 |
+
[
|
| 113 |
+
1.46318083,
|
| 114 |
+
0.16631192
|
| 115 |
+
],
|
| 116 |
+
[
|
| 117 |
+
2.79637467,
|
| 118 |
+
0.24332368
|
| 119 |
+
],
|
| 120 |
+
[
|
| 121 |
+
0.5,
|
| 122 |
+
0.5
|
| 123 |
+
],
|
| 124 |
+
[
|
| 125 |
+
1.12735104,
|
| 126 |
+
-0.11648428
|
| 127 |
+
],
|
| 128 |
+
[
|
| 129 |
+
1.45046443,
|
| 130 |
+
1.35436516
|
| 131 |
+
],
|
| 132 |
+
[
|
| 133 |
+
1.5324732,
|
| 134 |
+
1.45750941
|
| 135 |
+
],
|
| 136 |
+
[
|
| 137 |
+
1.80842297,
|
| 138 |
+
-0.01855904
|
| 139 |
+
],
|
| 140 |
+
[
|
| 141 |
+
1.46318083,
|
| 142 |
+
0.16631192
|
| 143 |
+
],
|
| 144 |
+
[
|
| 145 |
+
2.79637467,
|
| 146 |
+
0.24332368
|
| 147 |
+
],
|
| 148 |
+
[
|
| 149 |
+
0.5,
|
| 150 |
+
0.5
|
| 151 |
+
]
|
| 152 |
+
]
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ConvertDataType",
|
| 156 |
+
"convert_map": {
|
| 157 |
+
"imgs": "float32",
|
| 158 |
+
"depths": "float32",
|
| 159 |
+
"image_wh": "float32",
|
| 160 |
+
"projection_mat": "float32",
|
| 161 |
+
"embodiedment_mat": "float32"
|
| 162 |
+
}
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:DualArmKinematics",
|
| 166 |
+
"urdf": "./urdf/arx5_description_isaac.urdf"
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ItemSelection",
|
| 170 |
+
"keys": [
|
| 171 |
+
"imgs",
|
| 172 |
+
"depths",
|
| 173 |
+
"image_wh",
|
| 174 |
+
"projection_mat",
|
| 175 |
+
"embodiedment_mat",
|
| 176 |
+
"hist_robot_state",
|
| 177 |
+
"joint_scale_shift",
|
| 178 |
+
"kinematics",
|
| 179 |
+
"text",
|
| 180 |
+
"joint_mask"
|
| 181 |
+
]
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:UnsqueezeBatch"
|
| 185 |
+
}
|
| 186 |
+
]
|
| 187 |
+
},
|
| 188 |
+
"collate_fn": null,
|
| 189 |
+
"batch_size": 1
|
| 190 |
+
}
|
post_training_robotwin/robotwin2_0_ur5_wsg.config.json
ADDED
|
@@ -0,0 +1,227 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"load_depth": true,
|
| 10 |
+
"cam_names": [
|
| 11 |
+
"left_camera",
|
| 12 |
+
"right_camera",
|
| 13 |
+
"head_camera"
|
| 14 |
+
],
|
| 15 |
+
"valid_action_step": null,
|
| 16 |
+
"transforms": [
|
| 17 |
+
{
|
| 18 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddItems",
|
| 19 |
+
"T_base2world": [
|
| 20 |
+
[
|
| 21 |
+
1,
|
| 22 |
+
0,
|
| 23 |
+
0,
|
| 24 |
+
0
|
| 25 |
+
],
|
| 26 |
+
[
|
| 27 |
+
0,
|
| 28 |
+
1,
|
| 29 |
+
0,
|
| 30 |
+
-0.65
|
| 31 |
+
],
|
| 32 |
+
[
|
| 33 |
+
0,
|
| 34 |
+
0,
|
| 35 |
+
1,
|
| 36 |
+
0.65
|
| 37 |
+
],
|
| 38 |
+
[
|
| 39 |
+
0,
|
| 40 |
+
0,
|
| 41 |
+
0,
|
| 42 |
+
1
|
| 43 |
+
]
|
| 44 |
+
],
|
| 45 |
+
"joint_mask": [
|
| 46 |
+
true,
|
| 47 |
+
true,
|
| 48 |
+
true,
|
| 49 |
+
true,
|
| 50 |
+
true,
|
| 51 |
+
true,
|
| 52 |
+
false,
|
| 53 |
+
true,
|
| 54 |
+
true,
|
| 55 |
+
true,
|
| 56 |
+
true,
|
| 57 |
+
true,
|
| 58 |
+
true,
|
| 59 |
+
false
|
| 60 |
+
]
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:SimpleStateSampling",
|
| 64 |
+
"hist_steps": 1,
|
| 65 |
+
"pred_steps": 64
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:Resize",
|
| 69 |
+
"dst_wh": [
|
| 70 |
+
320,
|
| 71 |
+
256
|
| 72 |
+
]
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ImageChannelFlip",
|
| 76 |
+
"output_channel": [
|
| 77 |
+
2,
|
| 78 |
+
1,
|
| 79 |
+
0
|
| 80 |
+
]
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ToTensor"
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:MoveEgoToCam"
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:GetProjectionMat",
|
| 90 |
+
"target_coordinate": "ego"
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddScaleShift",
|
| 94 |
+
"scale_shift": [
|
| 95 |
+
[
|
| 96 |
+
2.400281548500061,
|
| 97 |
+
-0.1310516595840454
|
| 98 |
+
],
|
| 99 |
+
[
|
| 100 |
+
1.445511817932129,
|
| 101 |
+
-1.445511817932129
|
| 102 |
+
],
|
| 103 |
+
[
|
| 104 |
+
2.16847026348114,
|
| 105 |
+
-0.23492777347564697
|
| 106 |
+
],
|
| 107 |
+
[
|
| 108 |
+
1.7424615025520325,
|
| 109 |
+
-0.007538259029388428
|
| 110 |
+
],
|
| 111 |
+
[
|
| 112 |
+
2.8101450204849243,
|
| 113 |
+
0.15472495555877686
|
| 114 |
+
],
|
| 115 |
+
[
|
| 116 |
+
2.9653799533843994,
|
| 117 |
+
0.02583003044128418
|
| 118 |
+
],
|
| 119 |
+
[
|
| 120 |
+
0.5,
|
| 121 |
+
0.5
|
| 122 |
+
],
|
| 123 |
+
[
|
| 124 |
+
2.400281548500061,
|
| 125 |
+
-0.1310516595840454
|
| 126 |
+
],
|
| 127 |
+
[
|
| 128 |
+
1.445511817932129,
|
| 129 |
+
-1.445511817932129
|
| 130 |
+
],
|
| 131 |
+
[
|
| 132 |
+
2.16847026348114,
|
| 133 |
+
-0.23492777347564697
|
| 134 |
+
],
|
| 135 |
+
[
|
| 136 |
+
1.7424615025520325,
|
| 137 |
+
-0.007538259029388428
|
| 138 |
+
],
|
| 139 |
+
[
|
| 140 |
+
2.8101450204849243,
|
| 141 |
+
0.15472495555877686
|
| 142 |
+
],
|
| 143 |
+
[
|
| 144 |
+
2.9653799533843994,
|
| 145 |
+
0.02583003044128418
|
| 146 |
+
],
|
| 147 |
+
[
|
| 148 |
+
0.5,
|
| 149 |
+
0.5
|
| 150 |
+
]
|
| 151 |
+
]
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ConvertDataType",
|
| 155 |
+
"convert_map": {
|
| 156 |
+
"imgs": "float32",
|
| 157 |
+
"depths": "float32",
|
| 158 |
+
"image_wh": "float32",
|
| 159 |
+
"projection_mat": "float32",
|
| 160 |
+
"embodiedment_mat": "float32"
|
| 161 |
+
}
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:DualArmKinematics",
|
| 165 |
+
"urdf": "./urdf/robotwin2_dual_arm_ur5_wsg.urdf",
|
| 166 |
+
"left_arm_link_keys": [
|
| 167 |
+
"left_shoulder_link",
|
| 168 |
+
"left_upper_arm_link",
|
| 169 |
+
"left_forearm_link",
|
| 170 |
+
"left_wrist_1_link",
|
| 171 |
+
"left_wrist_2_link",
|
| 172 |
+
"left_wrist_3_link"
|
| 173 |
+
],
|
| 174 |
+
"left_finger_keys": [
|
| 175 |
+
"left_finger_left"
|
| 176 |
+
],
|
| 177 |
+
"right_arm_link_keys": [
|
| 178 |
+
"right_shoulder_link",
|
| 179 |
+
"right_upper_arm_link",
|
| 180 |
+
"right_forearm_link",
|
| 181 |
+
"right_wrist_1_link",
|
| 182 |
+
"right_wrist_2_link",
|
| 183 |
+
"right_wrist_3_link"
|
| 184 |
+
],
|
| 185 |
+
"right_finger_keys": [
|
| 186 |
+
"right_finger_left"
|
| 187 |
+
],
|
| 188 |
+
"left_arm_joint_id": [
|
| 189 |
+
0,
|
| 190 |
+
1,
|
| 191 |
+
2,
|
| 192 |
+
3,
|
| 193 |
+
4,
|
| 194 |
+
5
|
| 195 |
+
],
|
| 196 |
+
"right_arm_joint_id": [
|
| 197 |
+
8,
|
| 198 |
+
9,
|
| 199 |
+
10,
|
| 200 |
+
11,
|
| 201 |
+
12,
|
| 202 |
+
13
|
| 203 |
+
]
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ItemSelection",
|
| 207 |
+
"keys": [
|
| 208 |
+
"imgs",
|
| 209 |
+
"depths",
|
| 210 |
+
"image_wh",
|
| 211 |
+
"projection_mat",
|
| 212 |
+
"embodiedment_mat",
|
| 213 |
+
"hist_robot_state",
|
| 214 |
+
"joint_scale_shift",
|
| 215 |
+
"kinematics",
|
| 216 |
+
"text",
|
| 217 |
+
"joint_mask"
|
| 218 |
+
]
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:UnsqueezeBatch"
|
| 222 |
+
}
|
| 223 |
+
]
|
| 224 |
+
},
|
| 225 |
+
"collate_fn": null,
|
| 226 |
+
"batch_size": 1
|
| 227 |
+
}
|
pretrain/robotwin2_0.config.json
ADDED
|
@@ -0,0 +1,190 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"load_depth": true,
|
| 10 |
+
"cam_names": [
|
| 11 |
+
"front_camera",
|
| 12 |
+
"left_camera",
|
| 13 |
+
"right_camera",
|
| 14 |
+
"head_camera"
|
| 15 |
+
],
|
| 16 |
+
"valid_action_step": null,
|
| 17 |
+
"transforms": [
|
| 18 |
+
{
|
| 19 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddItems",
|
| 20 |
+
"T_base2world": [
|
| 21 |
+
[
|
| 22 |
+
0,
|
| 23 |
+
-1,
|
| 24 |
+
0,
|
| 25 |
+
0
|
| 26 |
+
],
|
| 27 |
+
[
|
| 28 |
+
1,
|
| 29 |
+
0,
|
| 30 |
+
0,
|
| 31 |
+
-0.65
|
| 32 |
+
],
|
| 33 |
+
[
|
| 34 |
+
0,
|
| 35 |
+
0,
|
| 36 |
+
1,
|
| 37 |
+
0
|
| 38 |
+
],
|
| 39 |
+
[
|
| 40 |
+
0,
|
| 41 |
+
0,
|
| 42 |
+
0,
|
| 43 |
+
1
|
| 44 |
+
]
|
| 45 |
+
],
|
| 46 |
+
"joint_mask": [
|
| 47 |
+
true,
|
| 48 |
+
true,
|
| 49 |
+
true,
|
| 50 |
+
true,
|
| 51 |
+
true,
|
| 52 |
+
true,
|
| 53 |
+
false,
|
| 54 |
+
true,
|
| 55 |
+
true,
|
| 56 |
+
true,
|
| 57 |
+
true,
|
| 58 |
+
true,
|
| 59 |
+
true,
|
| 60 |
+
false
|
| 61 |
+
]
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:SimpleStateSampling",
|
| 65 |
+
"hist_steps": 1,
|
| 66 |
+
"pred_steps": 64
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:Resize",
|
| 70 |
+
"dst_wh": [
|
| 71 |
+
320,
|
| 72 |
+
256
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ImageChannelFlip",
|
| 77 |
+
"output_channel": [
|
| 78 |
+
2,
|
| 79 |
+
1,
|
| 80 |
+
0
|
| 81 |
+
]
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ToTensor"
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:MoveEgoToCam"
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:GetProjectionMat",
|
| 91 |
+
"target_coordinate": "ego"
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddScaleShift",
|
| 95 |
+
"scale_shift": [
|
| 96 |
+
[
|
| 97 |
+
1.12735104,
|
| 98 |
+
-0.11648428
|
| 99 |
+
],
|
| 100 |
+
[
|
| 101 |
+
1.45046443,
|
| 102 |
+
1.35436516
|
| 103 |
+
],
|
| 104 |
+
[
|
| 105 |
+
1.5324732,
|
| 106 |
+
1.45750941
|
| 107 |
+
],
|
| 108 |
+
[
|
| 109 |
+
1.80842297,
|
| 110 |
+
-0.01855904
|
| 111 |
+
],
|
| 112 |
+
[
|
| 113 |
+
1.46318083,
|
| 114 |
+
0.16631192
|
| 115 |
+
],
|
| 116 |
+
[
|
| 117 |
+
2.79637467,
|
| 118 |
+
0.24332368
|
| 119 |
+
],
|
| 120 |
+
[
|
| 121 |
+
0.5,
|
| 122 |
+
0.5
|
| 123 |
+
],
|
| 124 |
+
[
|
| 125 |
+
1.12735104,
|
| 126 |
+
-0.11648428
|
| 127 |
+
],
|
| 128 |
+
[
|
| 129 |
+
1.45046443,
|
| 130 |
+
1.35436516
|
| 131 |
+
],
|
| 132 |
+
[
|
| 133 |
+
1.5324732,
|
| 134 |
+
1.45750941
|
| 135 |
+
],
|
| 136 |
+
[
|
| 137 |
+
1.80842297,
|
| 138 |
+
-0.01855904
|
| 139 |
+
],
|
| 140 |
+
[
|
| 141 |
+
1.46318083,
|
| 142 |
+
0.16631192
|
| 143 |
+
],
|
| 144 |
+
[
|
| 145 |
+
2.79637467,
|
| 146 |
+
0.24332368
|
| 147 |
+
],
|
| 148 |
+
[
|
| 149 |
+
0.5,
|
| 150 |
+
0.5
|
| 151 |
+
]
|
| 152 |
+
]
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ConvertDataType",
|
| 156 |
+
"convert_map": {
|
| 157 |
+
"imgs": "float32",
|
| 158 |
+
"depths": "float32",
|
| 159 |
+
"image_wh": "float32",
|
| 160 |
+
"projection_mat": "float32",
|
| 161 |
+
"embodiedment_mat": "float32"
|
| 162 |
+
}
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:DualArmKinematics",
|
| 166 |
+
"urdf": "./urdf/arx5_description_isaac.urdf"
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ItemSelection",
|
| 170 |
+
"keys": [
|
| 171 |
+
"imgs",
|
| 172 |
+
"depths",
|
| 173 |
+
"image_wh",
|
| 174 |
+
"projection_mat",
|
| 175 |
+
"embodiedment_mat",
|
| 176 |
+
"hist_robot_state",
|
| 177 |
+
"joint_scale_shift",
|
| 178 |
+
"kinematics",
|
| 179 |
+
"text",
|
| 180 |
+
"joint_mask"
|
| 181 |
+
]
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:UnsqueezeBatch"
|
| 185 |
+
}
|
| 186 |
+
]
|
| 187 |
+
},
|
| 188 |
+
"collate_fn": null,
|
| 189 |
+
"batch_size": 1
|
| 190 |
+
}
|
pretrain/robotwin2_0_ur5_wsg.config.json
ADDED
|
@@ -0,0 +1,227 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipelineCfg",
|
| 3 |
+
"class_type": "robo_orchard_lab.models.holobrain.pipeline:HoloBrainInferencePipeline",
|
| 4 |
+
"model_cfg": null,
|
| 5 |
+
"processor": {
|
| 6 |
+
"__config_type__": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessorCfg",
|
| 7 |
+
"class_type": "robo_orchard_lab.models.holobrain.processor:HoloBrainProcessor",
|
| 8 |
+
"load_image": true,
|
| 9 |
+
"load_depth": true,
|
| 10 |
+
"cam_names": [
|
| 11 |
+
"left_camera",
|
| 12 |
+
"right_camera",
|
| 13 |
+
"head_camera"
|
| 14 |
+
],
|
| 15 |
+
"valid_action_step": null,
|
| 16 |
+
"transforms": [
|
| 17 |
+
{
|
| 18 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddItems",
|
| 19 |
+
"T_base2world": [
|
| 20 |
+
[
|
| 21 |
+
1,
|
| 22 |
+
0,
|
| 23 |
+
0,
|
| 24 |
+
0
|
| 25 |
+
],
|
| 26 |
+
[
|
| 27 |
+
0,
|
| 28 |
+
1,
|
| 29 |
+
0,
|
| 30 |
+
-0.65
|
| 31 |
+
],
|
| 32 |
+
[
|
| 33 |
+
0,
|
| 34 |
+
0,
|
| 35 |
+
1,
|
| 36 |
+
0.65
|
| 37 |
+
],
|
| 38 |
+
[
|
| 39 |
+
0,
|
| 40 |
+
0,
|
| 41 |
+
0,
|
| 42 |
+
1
|
| 43 |
+
]
|
| 44 |
+
],
|
| 45 |
+
"joint_mask": [
|
| 46 |
+
true,
|
| 47 |
+
true,
|
| 48 |
+
true,
|
| 49 |
+
true,
|
| 50 |
+
true,
|
| 51 |
+
true,
|
| 52 |
+
false,
|
| 53 |
+
true,
|
| 54 |
+
true,
|
| 55 |
+
true,
|
| 56 |
+
true,
|
| 57 |
+
true,
|
| 58 |
+
true,
|
| 59 |
+
false
|
| 60 |
+
]
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:SimpleStateSampling",
|
| 64 |
+
"hist_steps": 1,
|
| 65 |
+
"pred_steps": 64
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:Resize",
|
| 69 |
+
"dst_wh": [
|
| 70 |
+
320,
|
| 71 |
+
256
|
| 72 |
+
]
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ImageChannelFlip",
|
| 76 |
+
"output_channel": [
|
| 77 |
+
2,
|
| 78 |
+
1,
|
| 79 |
+
0
|
| 80 |
+
]
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ToTensor"
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:MoveEgoToCam"
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:GetProjectionMat",
|
| 90 |
+
"target_coordinate": "ego"
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:AddScaleShift",
|
| 94 |
+
"scale_shift": [
|
| 95 |
+
[
|
| 96 |
+
2.400281548500061,
|
| 97 |
+
-0.1310516595840454
|
| 98 |
+
],
|
| 99 |
+
[
|
| 100 |
+
1.445511817932129,
|
| 101 |
+
-1.445511817932129
|
| 102 |
+
],
|
| 103 |
+
[
|
| 104 |
+
2.16847026348114,
|
| 105 |
+
-0.23492777347564697
|
| 106 |
+
],
|
| 107 |
+
[
|
| 108 |
+
1.7424615025520325,
|
| 109 |
+
-0.007538259029388428
|
| 110 |
+
],
|
| 111 |
+
[
|
| 112 |
+
2.8101450204849243,
|
| 113 |
+
0.15472495555877686
|
| 114 |
+
],
|
| 115 |
+
[
|
| 116 |
+
2.9653799533843994,
|
| 117 |
+
0.02583003044128418
|
| 118 |
+
],
|
| 119 |
+
[
|
| 120 |
+
0.5,
|
| 121 |
+
0.5
|
| 122 |
+
],
|
| 123 |
+
[
|
| 124 |
+
2.400281548500061,
|
| 125 |
+
-0.1310516595840454
|
| 126 |
+
],
|
| 127 |
+
[
|
| 128 |
+
1.445511817932129,
|
| 129 |
+
-1.445511817932129
|
| 130 |
+
],
|
| 131 |
+
[
|
| 132 |
+
2.16847026348114,
|
| 133 |
+
-0.23492777347564697
|
| 134 |
+
],
|
| 135 |
+
[
|
| 136 |
+
1.7424615025520325,
|
| 137 |
+
-0.007538259029388428
|
| 138 |
+
],
|
| 139 |
+
[
|
| 140 |
+
2.8101450204849243,
|
| 141 |
+
0.15472495555877686
|
| 142 |
+
],
|
| 143 |
+
[
|
| 144 |
+
2.9653799533843994,
|
| 145 |
+
0.02583003044128418
|
| 146 |
+
],
|
| 147 |
+
[
|
| 148 |
+
0.5,
|
| 149 |
+
0.5
|
| 150 |
+
]
|
| 151 |
+
]
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ConvertDataType",
|
| 155 |
+
"convert_map": {
|
| 156 |
+
"imgs": "float32",
|
| 157 |
+
"depths": "float32",
|
| 158 |
+
"image_wh": "float32",
|
| 159 |
+
"projection_mat": "float32",
|
| 160 |
+
"embodiedment_mat": "float32"
|
| 161 |
+
}
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:DualArmKinematics",
|
| 165 |
+
"urdf": "./urdf/robotwin2_dual_arm_ur5_wsg.urdf",
|
| 166 |
+
"left_arm_link_keys": [
|
| 167 |
+
"left_shoulder_link",
|
| 168 |
+
"left_upper_arm_link",
|
| 169 |
+
"left_forearm_link",
|
| 170 |
+
"left_wrist_1_link",
|
| 171 |
+
"left_wrist_2_link",
|
| 172 |
+
"left_wrist_3_link"
|
| 173 |
+
],
|
| 174 |
+
"left_finger_keys": [
|
| 175 |
+
"left_finger_left"
|
| 176 |
+
],
|
| 177 |
+
"right_arm_link_keys": [
|
| 178 |
+
"right_shoulder_link",
|
| 179 |
+
"right_upper_arm_link",
|
| 180 |
+
"right_forearm_link",
|
| 181 |
+
"right_wrist_1_link",
|
| 182 |
+
"right_wrist_2_link",
|
| 183 |
+
"right_wrist_3_link"
|
| 184 |
+
],
|
| 185 |
+
"right_finger_keys": [
|
| 186 |
+
"right_finger_left"
|
| 187 |
+
],
|
| 188 |
+
"left_arm_joint_id": [
|
| 189 |
+
0,
|
| 190 |
+
1,
|
| 191 |
+
2,
|
| 192 |
+
3,
|
| 193 |
+
4,
|
| 194 |
+
5
|
| 195 |
+
],
|
| 196 |
+
"right_arm_joint_id": [
|
| 197 |
+
8,
|
| 198 |
+
9,
|
| 199 |
+
10,
|
| 200 |
+
11,
|
| 201 |
+
12,
|
| 202 |
+
13
|
| 203 |
+
]
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:ItemSelection",
|
| 207 |
+
"keys": [
|
| 208 |
+
"imgs",
|
| 209 |
+
"depths",
|
| 210 |
+
"image_wh",
|
| 211 |
+
"projection_mat",
|
| 212 |
+
"embodiedment_mat",
|
| 213 |
+
"hist_robot_state",
|
| 214 |
+
"joint_scale_shift",
|
| 215 |
+
"kinematics",
|
| 216 |
+
"text",
|
| 217 |
+
"joint_mask"
|
| 218 |
+
]
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"type": "robo_orchard_lab.dataset.robotwin.transforms:UnsqueezeBatch"
|
| 222 |
+
}
|
| 223 |
+
]
|
| 224 |
+
},
|
| 225 |
+
"collate_fn": null,
|
| 226 |
+
"batch_size": 1
|
| 227 |
+
}
|