Hugging Face's logo

dreamdifferent
/
groot-n17-alps

Model card Files
xet
 Community
juliqchen commited on
Commit
3e4acae
·
verified ·
1 Parent(s): a1537dc

Publish source milestone 320000

Browse files
Files changed (9) hide show
  1. campaigns/02b73e57b7ee8f85/source/samples-320000/checkpoint.json +26 -0
  2. campaigns/02b73e57b7ee8f85/source/samples-320000/contract.json +71 -0
  3. campaigns/02b73e57b7ee8f85/source/samples-320000/processor/embodiment_id.json +57 -0
  4. campaigns/02b73e57b7ee8f85/source/samples-320000/processor/processor_config.json +1224 -0
  5. campaigns/02b73e57b7ee8f85/source/samples-320000/processor/statistics.json +0 -0
  6. campaigns/02b73e57b7ee8f85/source/samples-320000/run_config.json +103 -0
  7. campaigns/02b73e57b7ee8f85/source/samples-320000/trainable.pt +3 -0
  8. campaigns/02b73e57b7ee8f85/source/samples-320000/training.pt +3 -0
  9. campaigns/02b73e57b7ee8f85/source/samples-320000/validation_curve.json +302 -0
campaigns/02b73e57b7ee8f85/source/samples-320000/checkpoint.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 2,
3
+ "packet_sha256": "02b73e57b7ee8f857b66c7535d2a8ae93644a839eeab0952c9158f978eecf9b7",
4
+ "code_commit": "a1dfd917bcd4ba34a74f5727cfecede349f4027a",
5
+ "config_sha256": "f89abdc1be77d258e59a86986dcf1254c05a15eaff38aaa596106e6a38492757",
6
+ "run_id": "source",
7
+ "allocation_id": "8391d0479063447e82e4b4494c0f1b8c",
8
+ "parent_checkpoint": null,
9
+ "attempt_id": "a9bce14473004e87a0b1722ced4f1f32",
10
+ "complete": true,
11
+ "termination_reason": null,
12
+ "consumed_samples": 320000,
13
+ "milestones": [
14
+ 75968,
15
+ 320000
16
+ ],
17
+ "files": {
18
+ "validation_curve.json": 7160,
19
+ "contract.json": 2121,
20
+ "trainable.pt": 7581308653,
21
+ "training.pt": 15162844308,
22
+ "processor/processor_config.json": 28640,
23
+ "processor/statistics.json": 4318619,
24
+ "processor/embodiment_id.json": 2226
25
+ }
26
+ }
campaigns/02b73e57b7ee8f85/source/samples-320000/contract.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "condition": "source",
3
+ "weights": {
4
+ "model": {
5
+ "repo_id": "nvidia/GR00T-N1.7-SimplerEnv-Bridge",
6
+ "revision": "940134b3c2948ccfdf8e7393f2d2ca869dc42833",
7
+ "backbone_repo_id": "nvidia/Cosmos-Reason2-2B",
8
+ "backbone_revision": "9ce19a195e423419c349abfc86fd07178b230561",
9
+ "upstream_revision": "51d4c89f72fda44cbf77285c6a8114b52676b8a1",
10
+ "gradient_checkpointing": false,
11
+ "temporal_embodiment": "simpler_env_widowx"
12
+ },
13
+ "lora": {
14
+ "rank": 256,
15
+ "alpha": 32,
16
+ "dropout": 0.0,
17
+ "target_patterns": [
18
+ "backbone\\.model\\.model\\.visual\\.blocks\\.\\d+\\.(attn\\.(qkv|proj)|mlp\\.linear_fc[12])",
19
+ "backbone\\.model\\.model\\.visual\\.(deepstack_merger_list\\.\\d+|merger)\\.linear_fc[12]",
20
+ "backbone\\.model\\.model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.[qkvo]_proj|mlp\\.(gate|up|down)_proj)"
21
+ ]
22
+ },
23
+ "embodiment": "new_embodiment",
24
+ "cameras": [
25
+ "corner_cam",
26
+ "front_cam"
27
+ ],
28
+ "fps": 5,
29
+ "action_horizon": 8,
30
+ "padded_action_horizon": 40
31
+ },
32
+ "dataset_sha256": "04b560b54c3b50c03c5825b27ea26f49b333e69a60c57fdda1ac0ed045e8f991",
33
+ "processor": {
34
+ "embodiment_id.json": "c8af3d61fe9ca5508c4a5b9e8be6ad4d0d1eb2c68cec76620ea089b894a3ee22",
35
+ "processor_config.json": "79c660e928a93152bde75e539c9806957888f6d16f78aa8b4c18a7fec6dbb008",
36
+ "statistics.json": "7afbf0205607856c3c5f7aee4e15b5c22182a8595099bff315311e4ce9105df0"
37
+ },
38
+ "world_size": 4,
39
+ "training": {
40
+ "microbatch": 32,
41
+ "accumulation": 1,
42
+ "num_workers": 8,
43
+ "numeric_cache_episodes": 16,
44
+ "seed": 42,
45
+ "save_interval_seconds": 900,
46
+ "log_every_updates": 10,
47
+ "validation_batches": 2,
48
+ "periodic_validation": {
49
+ "every_updates": 50,
50
+ "samples": 64,
51
+ "seed": 42
52
+ },
53
+ "max_samples": 320000
54
+ },
55
+ "optimizer": {
56
+ "lora_lr": 0.00017782794100389227,
57
+ "action_lr": 0.0001,
58
+ "betas": [
59
+ 0.9,
60
+ 0.999
61
+ ],
62
+ "eps": 1e-08,
63
+ "weight_decay": 0.01,
64
+ "clip_norm": 1.0
65
+ },
66
+ "milestone_samples": [
67
+ 75968,
68
+ 320000
69
+ ],
70
+ "purpose": "train"
71
+ }
campaigns/02b73e57b7ee8f85/source/samples-320000/processor/embodiment_id.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "oxe_droid": 17,
4
+ "oxe_fractal": 18,
5
+ "oxe_language_table": 19,
6
+ "oxe_bridge": 20,
7
+ "unknown": 22,
8
+ "gr1_unified": 20,
9
+ "agibot": 26,
10
+ "sim_behavior_r1_pro": 23,
11
+ "xdof": 24,
12
+ "xdof_oss_data": 25,
13
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
14
+ "real_r1_pro_sharpa": 27,
15
+ "real_r1_pro_sharpa_add_view": 27,
16
+ "real_r1_pro_sharpa_relative_arm_joint": 26,
17
+ "real_r1_pro_sharpa_delta_eef": 26,
18
+ "real_r1_pro_sharpa_absolute_eef": 26,
19
+ "real_r1_pro_sharpa_meanstd": 26,
20
+ "real_r1_pro_sharpa_relative_eef": 26,
21
+ "real_r1_pro_sharpa_relative_eef_add_view": 26,
22
+ "real_r1_pro_sharpa_relative_eef_relative_hand": 26,
23
+ "real_r1_pro_sharpa_relative_eef_human": 26,
24
+ "real_r1_pro_sharpa_relative_eef_human_add_view": 26,
25
+ "real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
26
+ "real_r1_pro_sharpa_relative_eef_egodex": 26,
27
+ "real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
28
+ "real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
29
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
30
+ "real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
31
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
32
+ "real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
33
+ "real_g1_relative_eef_absolute_joints": 25,
34
+ "real_g1_relative_eef_absolute_joints_wrist_cam": 25,
35
+ "real_g1_relative_eef_relative_joints": 25,
36
+ "real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
37
+ "real_r1_pro_sharpa_relative_joint": 29,
38
+ "oxe_droid_relative_eef_relative_joint": 24,
39
+ "oxe_droid_relative_eef_relative_joint_swapped": 24,
40
+ "oxe_droid_relative_eef_relative_joint_upweight_z": 24,
41
+ "oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
42
+ "oxe_droid_relative_eef_relative_joint_3view": 24,
43
+ "oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
44
+ "oxe_droid_relative_eef": 24,
45
+ "oxe_droid_joint_position_relative": 24,
46
+ "xdof_relative_eef_relative_joint": 27,
47
+ "xdof_relative_eef_relative_joint_subtask": 27,
48
+ "xdof_relative_eef": 27,
49
+ "xdof_relative_joint": 28,
50
+ "simpler_env_google": 0,
51
+ "simpler_env_widowx": 1,
52
+ "libero_sim": 2,
53
+ "droid_sim": 3,
54
+ "new_embodiment": 10,
55
+ "robocasa_gr1_tabletop": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
campaigns/02b73e57b7ee8f85/source/samples-320000/processor/processor_config.json ADDED
@@ -0,0 +1,1224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "Gr00tN1d7Processor",
3
+ "processor_kwargs": {
4
+ "modality_configs": {
5
+ "real_g1_relative_eef_relative_joints": {
6
+ "video": {
7
+ "delta_indices": [
8
+ -20,
9
+ 0
10
+ ],
11
+ "modality_keys": [
12
+ "ego_view"
13
+ ],
14
+ "sin_cos_embedding_keys": null,
15
+ "mean_std_embedding_keys": null,
16
+ "action_configs": null
17
+ },
18
+ "state": {
19
+ "delta_indices": [
20
+ 0
21
+ ],
22
+ "modality_keys": [
23
+ "left_wrist_eef_9d",
24
+ "right_wrist_eef_9d",
25
+ "left_hand",
26
+ "right_hand",
27
+ "left_arm",
28
+ "right_arm",
29
+ "waist"
30
+ ],
31
+ "sin_cos_embedding_keys": null,
32
+ "mean_std_embedding_keys": null,
33
+ "action_configs": null
34
+ },
35
+ "action": {
36
+ "delta_indices": [
37
+ 0,
38
+ 1,
39
+ 2,
40
+ 3,
41
+ 4,
42
+ 5,
43
+ 6,
44
+ 7,
45
+ 8,
46
+ 9,
47
+ 10,
48
+ 11,
49
+ 12,
50
+ 13,
51
+ 14,
52
+ 15,
53
+ 16,
54
+ 17,
55
+ 18,
56
+ 19,
57
+ 20,
58
+ 21,
59
+ 22,
60
+ 23,
61
+ 24,
62
+ 25,
63
+ 26,
64
+ 27,
65
+ 28,
66
+ 29,
67
+ 30,
68
+ 31,
69
+ 32,
70
+ 33,
71
+ 34,
72
+ 35,
73
+ 36,
74
+ 37,
75
+ 38,
76
+ 39
77
+ ],
78
+ "modality_keys": [
79
+ "left_wrist_eef_9d",
80
+ "right_wrist_eef_9d",
81
+ "left_hand",
82
+ "right_hand",
83
+ "left_arm",
84
+ "right_arm",
85
+ "waist",
86
+ "base_height_command",
87
+ "navigate_command"
88
+ ],
89
+ "sin_cos_embedding_keys": null,
90
+ "mean_std_embedding_keys": null,
91
+ "action_configs": [
92
+ {
93
+ "rep": "RELATIVE",
94
+ "type": "EEF",
95
+ "format": "XYZ_ROT6D",
96
+ "state_key": null
97
+ },
98
+ {
99
+ "rep": "RELATIVE",
100
+ "type": "EEF",
101
+ "format": "XYZ_ROT6D",
102
+ "state_key": null
103
+ },
104
+ {
105
+ "rep": "ABSOLUTE",
106
+ "type": "NON_EEF",
107
+ "format": "DEFAULT",
108
+ "state_key": null
109
+ },
110
+ {
111
+ "rep": "ABSOLUTE",
112
+ "type": "NON_EEF",
113
+ "format": "DEFAULT",
114
+ "state_key": null
115
+ },
116
+ {
117
+ "rep": "RELATIVE",
118
+ "type": "NON_EEF",
119
+ "format": "DEFAULT",
120
+ "state_key": null
121
+ },
122
+ {
123
+ "rep": "RELATIVE",
124
+ "type": "NON_EEF",
125
+ "format": "DEFAULT",
126
+ "state_key": null
127
+ },
128
+ {
129
+ "rep": "ABSOLUTE",
130
+ "type": "NON_EEF",
131
+ "format": "DEFAULT",
132
+ "state_key": null
133
+ },
134
+ {
135
+ "rep": "ABSOLUTE",
136
+ "type": "NON_EEF",
137
+ "format": "DEFAULT",
138
+ "state_key": null
139
+ },
140
+ {
141
+ "rep": "ABSOLUTE",
142
+ "type": "NON_EEF",
143
+ "format": "DEFAULT",
144
+ "state_key": null
145
+ }
146
+ ]
147
+ },
148
+ "language": {
149
+ "delta_indices": [
150
+ 0
151
+ ],
152
+ "modality_keys": [
153
+ "annotation.human.task_description"
154
+ ],
155
+ "sin_cos_embedding_keys": null,
156
+ "mean_std_embedding_keys": null,
157
+ "action_configs": null
158
+ }
159
+ },
160
+ "real_r1_pro_sharpa_relative_eef_mecka": {
161
+ "video": {
162
+ "delta_indices": [
163
+ -30,
164
+ 0
165
+ ],
166
+ "modality_keys": [
167
+ "ego_view_cropratio_res320x240_freq30"
168
+ ],
169
+ "sin_cos_embedding_keys": null,
170
+ "mean_std_embedding_keys": null,
171
+ "action_configs": null
172
+ },
173
+ "state": {
174
+ "delta_indices": [
175
+ 0
176
+ ],
177
+ "modality_keys": [
178
+ "left_wrist_eef",
179
+ "right_wrist_eef",
180
+ "left_hand_joints",
181
+ "right_hand_joints"
182
+ ],
183
+ "sin_cos_embedding_keys": null,
184
+ "mean_std_embedding_keys": null,
185
+ "action_configs": null
186
+ },
187
+ "action": {
188
+ "delta_indices": [
189
+ 0,
190
+ 1,
191
+ 2,
192
+ 3,
193
+ 4,
194
+ 5,
195
+ 6,
196
+ 7,
197
+ 8,
198
+ 9,
199
+ 10,
200
+ 11,
201
+ 12,
202
+ 13,
203
+ 14,
204
+ 15,
205
+ 16,
206
+ 17,
207
+ 18,
208
+ 19,
209
+ 20,
210
+ 21,
211
+ 22,
212
+ 23,
213
+ 24,
214
+ 25,
215
+ 26,
216
+ 27,
217
+ 28,
218
+ 29,
219
+ 30,
220
+ 31,
221
+ 32,
222
+ 33,
223
+ 34,
224
+ 35,
225
+ 36,
226
+ 37,
227
+ 38,
228
+ 39
229
+ ],
230
+ "modality_keys": [
231
+ "left_wrist_eef",
232
+ "right_wrist_eef",
233
+ "left_hand_joints",
234
+ "right_hand_joints"
235
+ ],
236
+ "sin_cos_embedding_keys": null,
237
+ "mean_std_embedding_keys": null,
238
+ "action_configs": [
239
+ {
240
+ "rep": "RELATIVE",
241
+ "type": "EEF",
242
+ "format": "XYZ_ROT6D",
243
+ "state_key": null
244
+ },
245
+ {
246
+ "rep": "RELATIVE",
247
+ "type": "EEF",
248
+ "format": "XYZ_ROT6D",
249
+ "state_key": null
250
+ },
251
+ {
252
+ "rep": "ABSOLUTE",
253
+ "type": "NON_EEF",
254
+ "format": "DEFAULT",
255
+ "state_key": null
256
+ },
257
+ {
258
+ "rep": "ABSOLUTE",
259
+ "type": "NON_EEF",
260
+ "format": "DEFAULT",
261
+ "state_key": null
262
+ }
263
+ ]
264
+ },
265
+ "language": {
266
+ "delta_indices": [
267
+ 0
268
+ ],
269
+ "modality_keys": [
270
+ "annotation.human.coarse_action"
271
+ ],
272
+ "sin_cos_embedding_keys": null,
273
+ "mean_std_embedding_keys": null,
274
+ "action_configs": null
275
+ }
276
+ },
277
+ "real_r1_pro_sharpa_relative_eef_human": {
278
+ "video": {
279
+ "delta_indices": [
280
+ -20,
281
+ 0
282
+ ],
283
+ "modality_keys": [
284
+ "ego_view_res320x240_freq20",
285
+ "left_wrist_view_res320x240_freq20",
286
+ "right_wrist_view_res320x240_freq20"
287
+ ],
288
+ "sin_cos_embedding_keys": null,
289
+ "mean_std_embedding_keys": null,
290
+ "action_configs": null
291
+ },
292
+ "state": {
293
+ "delta_indices": [
294
+ 0
295
+ ],
296
+ "modality_keys": [
297
+ "left_wrist_eef",
298
+ "right_wrist_eef",
299
+ "left_hand_joints",
300
+ "right_hand_joints"
301
+ ],
302
+ "sin_cos_embedding_keys": null,
303
+ "mean_std_embedding_keys": null,
304
+ "action_configs": null
305
+ },
306
+ "action": {
307
+ "delta_indices": [
308
+ 0,
309
+ 1,
310
+ 2,
311
+ 3,
312
+ 4,
313
+ 5,
314
+ 6,
315
+ 7,
316
+ 8,
317
+ 9,
318
+ 10,
319
+ 11,
320
+ 12,
321
+ 13,
322
+ 14,
323
+ 15,
324
+ 16,
325
+ 17,
326
+ 18,
327
+ 19,
328
+ 20,
329
+ 21,
330
+ 22,
331
+ 23,
332
+ 24,
333
+ 25,
334
+ 26,
335
+ 27,
336
+ 28,
337
+ 29,
338
+ 30,
339
+ 31,
340
+ 32,
341
+ 33,
342
+ 34,
343
+ 35,
344
+ 36,
345
+ 37,
346
+ 38,
347
+ 39
348
+ ],
349
+ "modality_keys": [
350
+ "left_wrist_eef",
351
+ "right_wrist_eef",
352
+ "left_hand_joints",
353
+ "right_hand_joints"
354
+ ],
355
+ "sin_cos_embedding_keys": null,
356
+ "mean_std_embedding_keys": null,
357
+ "action_configs": [
358
+ {
359
+ "rep": "RELATIVE",
360
+ "type": "EEF",
361
+ "format": "XYZ_ROT6D",
362
+ "state_key": null
363
+ },
364
+ {
365
+ "rep": "RELATIVE",
366
+ "type": "EEF",
367
+ "format": "XYZ_ROT6D",
368
+ "state_key": null
369
+ },
370
+ {
371
+ "rep": "ABSOLUTE",
372
+ "type": "NON_EEF",
373
+ "format": "DEFAULT",
374
+ "state_key": null
375
+ },
376
+ {
377
+ "rep": "ABSOLUTE",
378
+ "type": "NON_EEF",
379
+ "format": "DEFAULT",
380
+ "state_key": null
381
+ }
382
+ ]
383
+ },
384
+ "language": {
385
+ "delta_indices": [
386
+ 0
387
+ ],
388
+ "modality_keys": [
389
+ "annotation.human.coarse_action"
390
+ ],
391
+ "sin_cos_embedding_keys": null,
392
+ "mean_std_embedding_keys": null,
393
+ "action_configs": null
394
+ }
395
+ },
396
+ "real_r1_pro_sharpa_relative_eef": {
397
+ "video": {
398
+ "delta_indices": [
399
+ -20,
400
+ 0
401
+ ],
402
+ "modality_keys": [
403
+ "ego_view_res320x240_freq20",
404
+ "left_wrist_view_res320x240_freq20",
405
+ "right_wrist_view_res320x240_freq20"
406
+ ],
407
+ "sin_cos_embedding_keys": null,
408
+ "mean_std_embedding_keys": null,
409
+ "action_configs": null
410
+ },
411
+ "state": {
412
+ "delta_indices": [
413
+ 0
414
+ ],
415
+ "modality_keys": [
416
+ "left_wrist_eef",
417
+ "right_wrist_eef",
418
+ "left_hand_joints",
419
+ "right_hand_joints"
420
+ ],
421
+ "sin_cos_embedding_keys": null,
422
+ "mean_std_embedding_keys": null,
423
+ "action_configs": null
424
+ },
425
+ "action": {
426
+ "delta_indices": [
427
+ 0,
428
+ 1,
429
+ 2,
430
+ 3,
431
+ 4,
432
+ 5,
433
+ 6,
434
+ 7,
435
+ 8,
436
+ 9,
437
+ 10,
438
+ 11,
439
+ 12,
440
+ 13,
441
+ 14,
442
+ 15,
443
+ 16,
444
+ 17,
445
+ 18,
446
+ 19,
447
+ 20,
448
+ 21,
449
+ 22,
450
+ 23,
451
+ 24,
452
+ 25,
453
+ 26,
454
+ 27,
455
+ 28,
456
+ 29,
457
+ 30,
458
+ 31,
459
+ 32,
460
+ 33,
461
+ 34,
462
+ 35,
463
+ 36,
464
+ 37,
465
+ 38,
466
+ 39
467
+ ],
468
+ "modality_keys": [
469
+ "left_wrist_eef",
470
+ "right_wrist_eef",
471
+ "left_hand_joints",
472
+ "right_hand_joints"
473
+ ],
474
+ "sin_cos_embedding_keys": null,
475
+ "mean_std_embedding_keys": null,
476
+ "action_configs": [
477
+ {
478
+ "rep": "RELATIVE",
479
+ "type": "EEF",
480
+ "format": "XYZ_ROT6D",
481
+ "state_key": null
482
+ },
483
+ {
484
+ "rep": "RELATIVE",
485
+ "type": "EEF",
486
+ "format": "XYZ_ROT6D",
487
+ "state_key": null
488
+ },
489
+ {
490
+ "rep": "ABSOLUTE",
491
+ "type": "NON_EEF",
492
+ "format": "DEFAULT",
493
+ "state_key": null
494
+ },
495
+ {
496
+ "rep": "ABSOLUTE",
497
+ "type": "NON_EEF",
498
+ "format": "DEFAULT",
499
+ "state_key": null
500
+ }
501
+ ]
502
+ },
503
+ "language": {
504
+ "delta_indices": [
505
+ 0
506
+ ],
507
+ "modality_keys": [
508
+ "annotation.human.coarse_action"
509
+ ],
510
+ "sin_cos_embedding_keys": null,
511
+ "mean_std_embedding_keys": null,
512
+ "action_configs": null
513
+ }
514
+ },
515
+ "xdof_relative_eef_relative_joint": {
516
+ "video": {
517
+ "delta_indices": [
518
+ -30,
519
+ 0
520
+ ],
521
+ "modality_keys": [
522
+ "top_camera-images-rgb_320_240",
523
+ "left_camera-images-rgb_320_240",
524
+ "right_camera-images-rgb_320_240"
525
+ ],
526
+ "sin_cos_embedding_keys": null,
527
+ "mean_std_embedding_keys": null,
528
+ "action_configs": null
529
+ },
530
+ "state": {
531
+ "delta_indices": [
532
+ 0
533
+ ],
534
+ "modality_keys": [
535
+ "left_wrist_eef",
536
+ "right_wrist_eef",
537
+ "left_gripper_pos",
538
+ "right_gripper_pos",
539
+ "left_joint_pos",
540
+ "right_joint_pos"
541
+ ],
542
+ "sin_cos_embedding_keys": null,
543
+ "mean_std_embedding_keys": null,
544
+ "action_configs": null
545
+ },
546
+ "action": {
547
+ "delta_indices": [
548
+ 0,
549
+ 1,
550
+ 2,
551
+ 3,
552
+ 4,
553
+ 5,
554
+ 6,
555
+ 7,
556
+ 8,
557
+ 9,
558
+ 10,
559
+ 11,
560
+ 12,
561
+ 13,
562
+ 14,
563
+ 15,
564
+ 16,
565
+ 17,
566
+ 18,
567
+ 19,
568
+ 20,
569
+ 21,
570
+ 22,
571
+ 23,
572
+ 24,
573
+ 25,
574
+ 26,
575
+ 27,
576
+ 28,
577
+ 29,
578
+ 30,
579
+ 31,
580
+ 32,
581
+ 33,
582
+ 34,
583
+ 35,
584
+ 36,
585
+ 37,
586
+ 38,
587
+ 39
588
+ ],
589
+ "modality_keys": [
590
+ "left_wrist_eef",
591
+ "right_wrist_eef",
592
+ "left_gripper_pos",
593
+ "right_gripper_pos",
594
+ "left_joint_pos",
595
+ "right_joint_pos"
596
+ ],
597
+ "sin_cos_embedding_keys": null,
598
+ "mean_std_embedding_keys": null,
599
+ "action_configs": [
600
+ {
601
+ "rep": "RELATIVE",
602
+ "type": "EEF",
603
+ "format": "XYZ_ROT6D",
604
+ "state_key": null
605
+ },
606
+ {
607
+ "rep": "RELATIVE",
608
+ "type": "EEF",
609
+ "format": "XYZ_ROT6D",
610
+ "state_key": null
611
+ },
612
+ {
613
+ "rep": "ABSOLUTE",
614
+ "type": "NON_EEF",
615
+ "format": "DEFAULT",
616
+ "state_key": null
617
+ },
618
+ {
619
+ "rep": "ABSOLUTE",
620
+ "type": "NON_EEF",
621
+ "format": "DEFAULT",
622
+ "state_key": null
623
+ },
624
+ {
625
+ "rep": "RELATIVE",
626
+ "type": "NON_EEF",
627
+ "format": "DEFAULT",
628
+ "state_key": null
629
+ },
630
+ {
631
+ "rep": "RELATIVE",
632
+ "type": "NON_EEF",
633
+ "format": "DEFAULT",
634
+ "state_key": null
635
+ }
636
+ ]
637
+ },
638
+ "language": {
639
+ "delta_indices": [
640
+ 0
641
+ ],
642
+ "modality_keys": [
643
+ "annotation.task"
644
+ ],
645
+ "sin_cos_embedding_keys": null,
646
+ "mean_std_embedding_keys": null,
647
+ "action_configs": null
648
+ }
649
+ },
650
+ "real_r1_pro_sharpa_relative_eef_maxinsights": {
651
+ "video": {
652
+ "delta_indices": [
653
+ -30,
654
+ 0
655
+ ],
656
+ "modality_keys": [
657
+ "ego_view_cropratio_res320x240_freq30"
658
+ ],
659
+ "sin_cos_embedding_keys": null,
660
+ "mean_std_embedding_keys": null,
661
+ "action_configs": null
662
+ },
663
+ "state": {
664
+ "delta_indices": [
665
+ 0
666
+ ],
667
+ "modality_keys": [
668
+ "left_wrist_eef",
669
+ "right_wrist_eef",
670
+ "left_hand_joints",
671
+ "right_hand_joints"
672
+ ],
673
+ "sin_cos_embedding_keys": null,
674
+ "mean_std_embedding_keys": null,
675
+ "action_configs": null
676
+ },
677
+ "action": {
678
+ "delta_indices": [
679
+ 0,
680
+ 1,
681
+ 2,
682
+ 3,
683
+ 4,
684
+ 5,
685
+ 6,
686
+ 7,
687
+ 8,
688
+ 9,
689
+ 10,
690
+ 11,
691
+ 12,
692
+ 13,
693
+ 14,
694
+ 15,
695
+ 16,
696
+ 17,
697
+ 18,
698
+ 19,
699
+ 20,
700
+ 21,
701
+ 22,
702
+ 23,
703
+ 24,
704
+ 25,
705
+ 26,
706
+ 27,
707
+ 28,
708
+ 29,
709
+ 30,
710
+ 31,
711
+ 32,
712
+ 33,
713
+ 34,
714
+ 35,
715
+ 36,
716
+ 37,
717
+ 38,
718
+ 39
719
+ ],
720
+ "modality_keys": [
721
+ "left_wrist_eef",
722
+ "right_wrist_eef",
723
+ "left_hand_joints",
724
+ "right_hand_joints"
725
+ ],
726
+ "sin_cos_embedding_keys": null,
727
+ "mean_std_embedding_keys": null,
728
+ "action_configs": [
729
+ {
730
+ "rep": "RELATIVE",
731
+ "type": "EEF",
732
+ "format": "XYZ_ROT6D",
733
+ "state_key": null
734
+ },
735
+ {
736
+ "rep": "RELATIVE",
737
+ "type": "EEF",
738
+ "format": "XYZ_ROT6D",
739
+ "state_key": null
740
+ },
741
+ {
742
+ "rep": "ABSOLUTE",
743
+ "type": "NON_EEF",
744
+ "format": "DEFAULT",
745
+ "state_key": null
746
+ },
747
+ {
748
+ "rep": "ABSOLUTE",
749
+ "type": "NON_EEF",
750
+ "format": "DEFAULT",
751
+ "state_key": null
752
+ }
753
+ ]
754
+ },
755
+ "language": {
756
+ "delta_indices": [
757
+ 0
758
+ ],
759
+ "modality_keys": [
760
+ "annotation.human.coarse_action"
761
+ ],
762
+ "sin_cos_embedding_keys": null,
763
+ "mean_std_embedding_keys": null,
764
+ "action_configs": null
765
+ }
766
+ },
767
+ "xdof_relative_eef_relative_joint_subtask": {
768
+ "video": {
769
+ "delta_indices": [
770
+ -30,
771
+ 0
772
+ ],
773
+ "modality_keys": [
774
+ "top_camera-images-rgb_320_240",
775
+ "left_camera-images-rgb_320_240",
776
+ "right_camera-images-rgb_320_240"
777
+ ],
778
+ "sin_cos_embedding_keys": null,
779
+ "mean_std_embedding_keys": null,
780
+ "action_configs": null
781
+ },
782
+ "state": {
783
+ "delta_indices": [
784
+ 0
785
+ ],
786
+ "modality_keys": [
787
+ "left_wrist_eef",
788
+ "right_wrist_eef",
789
+ "left_gripper_pos",
790
+ "right_gripper_pos",
791
+ "left_joint_pos",
792
+ "right_joint_pos"
793
+ ],
794
+ "sin_cos_embedding_keys": null,
795
+ "mean_std_embedding_keys": null,
796
+ "action_configs": null
797
+ },
798
+ "action": {
799
+ "delta_indices": [
800
+ 0,
801
+ 1,
802
+ 2,
803
+ 3,
804
+ 4,
805
+ 5,
806
+ 6,
807
+ 7,
808
+ 8,
809
+ 9,
810
+ 10,
811
+ 11,
812
+ 12,
813
+ 13,
814
+ 14,
815
+ 15,
816
+ 16,
817
+ 17,
818
+ 18,
819
+ 19,
820
+ 20,
821
+ 21,
822
+ 22,
823
+ 23,
824
+ 24,
825
+ 25,
826
+ 26,
827
+ 27,
828
+ 28,
829
+ 29,
830
+ 30,
831
+ 31,
832
+ 32,
833
+ 33,
834
+ 34,
835
+ 35,
836
+ 36,
837
+ 37,
838
+ 38,
839
+ 39
840
+ ],
841
+ "modality_keys": [
842
+ "left_wrist_eef",
843
+ "right_wrist_eef",
844
+ "left_gripper_pos",
845
+ "right_gripper_pos",
846
+ "left_joint_pos",
847
+ "right_joint_pos"
848
+ ],
849
+ "sin_cos_embedding_keys": null,
850
+ "mean_std_embedding_keys": null,
851
+ "action_configs": [
852
+ {
853
+ "rep": "RELATIVE",
854
+ "type": "EEF",
855
+ "format": "XYZ_ROT6D",
856
+ "state_key": null
857
+ },
858
+ {
859
+ "rep": "RELATIVE",
860
+ "type": "EEF",
861
+ "format": "XYZ_ROT6D",
862
+ "state_key": null
863
+ },
864
+ {
865
+ "rep": "ABSOLUTE",
866
+ "type": "NON_EEF",
867
+ "format": "DEFAULT",
868
+ "state_key": null
869
+ },
870
+ {
871
+ "rep": "ABSOLUTE",
872
+ "type": "NON_EEF",
873
+ "format": "DEFAULT",
874
+ "state_key": null
875
+ },
876
+ {
877
+ "rep": "RELATIVE",
878
+ "type": "NON_EEF",
879
+ "format": "DEFAULT",
880
+ "state_key": null
881
+ },
882
+ {
883
+ "rep": "RELATIVE",
884
+ "type": "NON_EEF",
885
+ "format": "DEFAULT",
886
+ "state_key": null
887
+ }
888
+ ]
889
+ },
890
+ "language": {
891
+ "delta_indices": [
892
+ 0
893
+ ],
894
+ "modality_keys": [
895
+ "annotation.sub_task"
896
+ ],
897
+ "sin_cos_embedding_keys": null,
898
+ "mean_std_embedding_keys": null,
899
+ "action_configs": null
900
+ }
901
+ },
902
+ "oxe_droid_relative_eef_relative_joint": {
903
+ "video": {
904
+ "delta_indices": [
905
+ -15,
906
+ 0
907
+ ],
908
+ "modality_keys": [
909
+ "exterior_image_1_left",
910
+ "wrist_image_left"
911
+ ],
912
+ "sin_cos_embedding_keys": null,
913
+ "mean_std_embedding_keys": null,
914
+ "action_configs": null
915
+ },
916
+ "state": {
917
+ "delta_indices": [
918
+ 0
919
+ ],
920
+ "modality_keys": [
921
+ "eef_9d",
922
+ "gripper_position",
923
+ "joint_position"
924
+ ],
925
+ "sin_cos_embedding_keys": null,
926
+ "mean_std_embedding_keys": null,
927
+ "action_configs": null
928
+ },
929
+ "action": {
930
+ "delta_indices": [
931
+ 0,
932
+ 1,
933
+ 2,
934
+ 3,
935
+ 4,
936
+ 5,
937
+ 6,
938
+ 7,
939
+ 8,
940
+ 9,
941
+ 10,
942
+ 11,
943
+ 12,
944
+ 13,
945
+ 14,
946
+ 15,
947
+ 16,
948
+ 17,
949
+ 18,
950
+ 19,
951
+ 20,
952
+ 21,
953
+ 22,
954
+ 23,
955
+ 24,
956
+ 25,
957
+ 26,
958
+ 27,
959
+ 28,
960
+ 29,
961
+ 30,
962
+ 31,
963
+ 32,
964
+ 33,
965
+ 34,
966
+ 35,
967
+ 36,
968
+ 37,
969
+ 38,
970
+ 39
971
+ ],
972
+ "modality_keys": [
973
+ "eef_9d",
974
+ "gripper_position",
975
+ "joint_position"
976
+ ],
977
+ "sin_cos_embedding_keys": null,
978
+ "mean_std_embedding_keys": null,
979
+ "action_configs": [
980
+ {
981
+ "rep": "RELATIVE",
982
+ "type": "EEF",
983
+ "format": "XYZ_ROT6D",
984
+ "state_key": null
985
+ },
986
+ {
987
+ "rep": "ABSOLUTE",
988
+ "type": "NON_EEF",
989
+ "format": "DEFAULT",
990
+ "state_key": null
991
+ },
992
+ {
993
+ "rep": "RELATIVE",
994
+ "type": "NON_EEF",
995
+ "format": "DEFAULT",
996
+ "state_key": null
997
+ }
998
+ ]
999
+ },
1000
+ "language": {
1001
+ "delta_indices": [
1002
+ 0
1003
+ ],
1004
+ "modality_keys": [
1005
+ "annotation.language.language_instruction"
1006
+ ],
1007
+ "sin_cos_embedding_keys": null,
1008
+ "mean_std_embedding_keys": null,
1009
+ "action_configs": null
1010
+ }
1011
+ },
1012
+ "simpler_env_widowx": {
1013
+ "video": {
1014
+ "delta_indices": [
1015
+ 0
1016
+ ],
1017
+ "modality_keys": [
1018
+ "image_0"
1019
+ ],
1020
+ "sin_cos_embedding_keys": null,
1021
+ "mean_std_embedding_keys": null,
1022
+ "action_configs": null
1023
+ },
1024
+ "state": {
1025
+ "delta_indices": [
1026
+ 0
1027
+ ],
1028
+ "modality_keys": [
1029
+ "x",
1030
+ "y",
1031
+ "z",
1032
+ "roll",
1033
+ "pitch",
1034
+ "yaw",
1035
+ "pad",
1036
+ "gripper"
1037
+ ],
1038
+ "sin_cos_embedding_keys": null,
1039
+ "mean_std_embedding_keys": null,
1040
+ "action_configs": null
1041
+ },
1042
+ "action": {
1043
+ "delta_indices": [
1044
+ 0,
1045
+ 1,
1046
+ 2,
1047
+ 3,
1048
+ 4,
1049
+ 5,
1050
+ 6,
1051
+ 7
1052
+ ],
1053
+ "modality_keys": [
1054
+ "x",
1055
+ "y",
1056
+ "z",
1057
+ "roll",
1058
+ "pitch",
1059
+ "yaw",
1060
+ "gripper"
1061
+ ],
1062
+ "sin_cos_embedding_keys": null,
1063
+ "mean_std_embedding_keys": null,
1064
+ "action_configs": [
1065
+ {
1066
+ "rep": "ABSOLUTE",
1067
+ "type": "NON_EEF",
1068
+ "format": "DEFAULT",
1069
+ "state_key": null
1070
+ },
1071
+ {
1072
+ "rep": "ABSOLUTE",
1073
+ "type": "NON_EEF",
1074
+ "format": "DEFAULT",
1075
+ "state_key": null
1076
+ },
1077
+ {
1078
+ "rep": "ABSOLUTE",
1079
+ "type": "NON_EEF",
1080
+ "format": "DEFAULT",
1081
+ "state_key": null
1082
+ },
1083
+ {
1084
+ "rep": "ABSOLUTE",
1085
+ "type": "NON_EEF",
1086
+ "format": "DEFAULT",
1087
+ "state_key": null
1088
+ },
1089
+ {
1090
+ "rep": "ABSOLUTE",
1091
+ "type": "NON_EEF",
1092
+ "format": "DEFAULT",
1093
+ "state_key": null
1094
+ },
1095
+ {
1096
+ "rep": "ABSOLUTE",
1097
+ "type": "NON_EEF",
1098
+ "format": "DEFAULT",
1099
+ "state_key": null
1100
+ },
1101
+ {
1102
+ "rep": "ABSOLUTE",
1103
+ "type": "NON_EEF",
1104
+ "format": "DEFAULT",
1105
+ "state_key": null
1106
+ }
1107
+ ]
1108
+ },
1109
+ "language": {
1110
+ "delta_indices": [
1111
+ 0
1112
+ ],
1113
+ "modality_keys": [
1114
+ "annotation.human.action.task_description"
1115
+ ],
1116
+ "sin_cos_embedding_keys": null,
1117
+ "mean_std_embedding_keys": null,
1118
+ "action_configs": null
1119
+ }
1120
+ },
1121
+ "new_embodiment": {
1122
+ "video": {
1123
+ "delta_indices": [
1124
+ 0
1125
+ ],
1126
+ "modality_keys": [
1127
+ "corner_cam",
1128
+ "front_cam"
1129
+ ],
1130
+ "sin_cos_embedding_keys": null,
1131
+ "mean_std_embedding_keys": null,
1132
+ "action_configs": null
1133
+ },
1134
+ "state": {
1135
+ "delta_indices": [
1136
+ 0
1137
+ ],
1138
+ "modality_keys": [
1139
+ "eef",
1140
+ "gripper"
1141
+ ],
1142
+ "sin_cos_embedding_keys": null,
1143
+ "mean_std_embedding_keys": null,
1144
+ "action_configs": null
1145
+ },
1146
+ "action": {
1147
+ "delta_indices": [
1148
+ 0,
1149
+ 1,
1150
+ 2,
1151
+ 3,
1152
+ 4,
1153
+ 5,
1154
+ 6,
1155
+ 7
1156
+ ],
1157
+ "modality_keys": [
1158
+ "eef",
1159
+ "gripper"
1160
+ ],
1161
+ "sin_cos_embedding_keys": null,
1162
+ "mean_std_embedding_keys": null,
1163
+ "action_configs": [
1164
+ {
1165
+ "rep": "RELATIVE",
1166
+ "type": "EEF",
1167
+ "format": "XYZ_ROT6D",
1168
+ "state_key": "eef"
1169
+ },
1170
+ {
1171
+ "rep": "ABSOLUTE",
1172
+ "type": "NON_EEF",
1173
+ "format": "DEFAULT",
1174
+ "state_key": null
1175
+ }
1176
+ ]
1177
+ },
1178
+ "language": {
1179
+ "delta_indices": [
1180
+ 0
1181
+ ],
1182
+ "modality_keys": [
1183
+ "annotation.human.task_description"
1184
+ ],
1185
+ "sin_cos_embedding_keys": null,
1186
+ "mean_std_embedding_keys": null,
1187
+ "action_configs": null
1188
+ }
1189
+ }
1190
+ },
1191
+ "image_crop_size": [
1192
+ 230,
1193
+ 230
1194
+ ],
1195
+ "image_target_size": [
1196
+ 256,
1197
+ 256
1198
+ ],
1199
+ "use_albumentations": true,
1200
+ "random_rotation_angle": 0,
1201
+ "color_jitter_params": {
1202
+ "brightness": 0.3,
1203
+ "contrast": 0.4,
1204
+ "saturation": 0.5,
1205
+ "hue": 0.08
1206
+ },
1207
+ "shortest_image_edge": 256,
1208
+ "crop_fraction": 0.95,
1209
+ "letter_box_transform": false,
1210
+ "model_name": "/iopsstor/scratch/cscs/jiaqchen/groot_n17/campaign/production-bridge-s3-configs/model-inputs/nvidia/Cosmos-Reason2-2B",
1211
+ "model_type": "qwen",
1212
+ "formalize_language": true,
1213
+ "max_state_dim": 132,
1214
+ "max_action_dim": 132,
1215
+ "max_action_horizon": 40,
1216
+ "use_percentiles": true,
1217
+ "use_mean_std": false,
1218
+ "clip_outliers": true,
1219
+ "apply_sincos_state_encoding": false,
1220
+ "use_relative_action": true,
1221
+ "exclude_state": false,
1222
+ "state_dropout_prob": 0.0
1223
+ }
1224
+ }
campaigns/02b73e57b7ee8f85/source/samples-320000/processor/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
campaigns/02b73e57b7ee8f85/source/samples-320000/run_config.json ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 8,
3
+ "budget": {
4
+ "epochs": [
5
+ 8.994553634856738
6
+ ],
7
+ "milestone_samples": [
8
+ 75968
9
+ ],
10
+ "reference_train_anchors": 8446,
11
+ "train_anchors": 8446
12
+ },
13
+ "cameras": [
14
+ "corner_cam",
15
+ "front_cam"
16
+ ],
17
+ "condition": "source",
18
+ "dataset": {
19
+ "repo_id": "dreamdifferent/vam-cross-target-widowx250-native",
20
+ "revision": "44c28a3bada463f06845daee8a1dd87607d32956",
21
+ "robot": "widowx250"
22
+ },
23
+ "embodiment": "new_embodiment",
24
+ "export_path": null,
25
+ "fps": 5,
26
+ "frame": "native_grip_site_v1",
27
+ "initialize_from": null,
28
+ "lora": {
29
+ "alpha": 32,
30
+ "dropout": 0.0,
31
+ "rank": 256,
32
+ "target_patterns": [
33
+ "backbone\\.model\\.model\\.visual\\.blocks\\.\\d+\\.(attn\\.(qkv|proj)|mlp\\.linear_fc[12])",
34
+ "backbone\\.model\\.model\\.visual\\.(deepstack_merger_list\\.\\d+|merger)\\.linear_fc[12]",
35
+ "backbone\\.model\\.model\\.language_model\\.layers\\.\\d+\\.(self_attn\\.[qkvo]_proj|mlp\\.(gate|up|down)_proj)"
36
+ ]
37
+ },
38
+ "milestone_samples": [
39
+ 75968,
40
+ 320000
41
+ ],
42
+ "model": {
43
+ "backbone_path": null,
44
+ "backbone_repo_id": "nvidia/Cosmos-Reason2-2B",
45
+ "backbone_revision": "9ce19a195e423419c349abfc86fd07178b230561",
46
+ "gradient_checkpointing": false,
47
+ "path": null,
48
+ "repo_id": "nvidia/GR00T-N1.7-SimplerEnv-Bridge",
49
+ "revision": "940134b3c2948ccfdf8e7393f2d2ca869dc42833",
50
+ "temporal_embodiment": "simpler_env_widowx",
51
+ "upstream_path": null,
52
+ "upstream_revision": "51d4c89f72fda44cbf77285c6a8114b52676b8a1"
53
+ },
54
+ "optimizer": {
55
+ "action_lr": 0.0001,
56
+ "betas": [
57
+ 0.9,
58
+ 0.999
59
+ ],
60
+ "clip_norm": 1.0,
61
+ "eps": 1e-08,
62
+ "lora_lr": 0.00017782794100389227,
63
+ "weight_decay": 0.01
64
+ },
65
+ "padded_action_horizon": 40,
66
+ "purpose": "train",
67
+ "readiness": {
68
+ "initial_save_margin_seconds": 600,
69
+ "microbatch_candidates": [
70
+ 4,
71
+ 8,
72
+ 16,
73
+ 32,
74
+ 64
75
+ ]
76
+ },
77
+ "reference_batch": 32,
78
+ "reference_config": "configs/mimicvideo/widowx_target_w2a.json",
79
+ "reference_fps": 5,
80
+ "reference_updates": [
81
+ 2374
82
+ ],
83
+ "split_seed": 42,
84
+ "training": {
85
+ "accumulation": 1,
86
+ "log_every_updates": 10,
87
+ "max_samples": 320000,
88
+ "microbatch": 32,
89
+ "num_workers": 8,
90
+ "numeric_cache_episodes": 16,
91
+ "periodic_validation": {
92
+ "every_updates": 50,
93
+ "samples": 64,
94
+ "seed": 42
95
+ },
96
+ "save_interval_seconds": 900,
97
+ "seed": 42,
98
+ "validation_batches": 2
99
+ },
100
+ "validation_count": 10,
101
+ "video_crf": 18,
102
+ "world_size": 4
103
+ }
campaigns/02b73e57b7ee8f85/source/samples-320000/trainable.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c52bd0063fdf076dd5b49553c78b9e338156e860272edeac9574965a15084df
3
+ size 7581308653
campaigns/02b73e57b7ee8f85/source/samples-320000/training.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8a46f2cadf9980b07f9ab61e69441e0b77182a5628b65e0f60fa211811fcf1f
3
+ size 15162844308
campaigns/02b73e57b7ee8f85/source/samples-320000/validation_curve.json ADDED
@@ -0,0 +1,302 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "updates": 50,
4
+ "consumed_samples": 6400,
5
+ "train_loss": 0.9535726472735405,
6
+ "validation_loss": 0.6688453196547925
7
+ },
8
+ {
9
+ "updates": 100,
10
+ "consumed_samples": 12800,
11
+ "train_loss": 0.5262131620943546,
12
+ "validation_loss": 0.42265566368587315
13
+ },
14
+ {
15
+ "updates": 150,
16
+ "consumed_samples": 19200,
17
+ "train_loss": 0.3944579327106476,
18
+ "validation_loss": 0.35885322326794267
19
+ },
20
+ {
21
+ "updates": 200,
22
+ "consumed_samples": 25600,
23
+ "train_loss": 0.22082673702389002,
24
+ "validation_loss": 0.11888543429085985
25
+ },
26
+ {
27
+ "updates": 250,
28
+ "consumed_samples": 32000,
29
+ "train_loss": 0.11731862869113684,
30
+ "validation_loss": 0.09881346579641104
31
+ },
32
+ {
33
+ "updates": 300,
34
+ "consumed_samples": 38400,
35
+ "train_loss": 0.10749875292181969,
36
+ "validation_loss": 0.1004789058060851
37
+ },
38
+ {
39
+ "updates": 350,
40
+ "consumed_samples": 44800,
41
+ "train_loss": 0.09997334372252226,
42
+ "validation_loss": 0.09542400010104757
43
+ },
44
+ {
45
+ "updates": 400,
46
+ "consumed_samples": 51200,
47
+ "train_loss": 0.09362963806837797,
48
+ "validation_loss": 0.08914007782004774
49
+ },
50
+ {
51
+ "updates": 450,
52
+ "consumed_samples": 57600,
53
+ "train_loss": 0.0875530356541276,
54
+ "validation_loss": 0.08726009902602527
55
+ },
56
+ {
57
+ "updates": 500,
58
+ "consumed_samples": 64000,
59
+ "train_loss": 0.08596241874620318,
60
+ "validation_loss": 0.09621652433997951
61
+ },
62
+ {
63
+ "updates": 550,
64
+ "consumed_samples": 70400,
65
+ "train_loss": 0.0821264905296266,
66
+ "validation_loss": 0.08541787066496909
67
+ },
68
+ {
69
+ "updates": 600,
70
+ "consumed_samples": 76800,
71
+ "train_loss": 0.07841321865096688,
72
+ "validation_loss": 0.09395434215548448
73
+ },
74
+ {
75
+ "updates": 650,
76
+ "consumed_samples": 83200,
77
+ "train_loss": 0.07812179919332266,
78
+ "validation_loss": 0.0882736288185697
79
+ },
80
+ {
81
+ "updates": 700,
82
+ "consumed_samples": 89600,
83
+ "train_loss": 0.07474292609840631,
84
+ "validation_loss": 0.08291742837900529
85
+ },
86
+ {
87
+ "updates": 750,
88
+ "consumed_samples": 96000,
89
+ "train_loss": 0.07077176762744784,
90
+ "validation_loss": 0.08862894825870171
91
+ },
92
+ {
93
+ "updates": 800,
94
+ "consumed_samples": 102400,
95
+ "train_loss": 0.06893341036513448,
96
+ "validation_loss": 0.09006108345784014
97
+ },
98
+ {
99
+ "updates": 850,
100
+ "consumed_samples": 108800,
101
+ "train_loss": 0.06813665878027678,
102
+ "validation_loss": 0.08452020486583933
103
+ },
104
+ {
105
+ "updates": 900,
106
+ "consumed_samples": 115200,
107
+ "train_loss": 0.0660622571222484,
108
+ "validation_loss": 0.09812473282363499
109
+ },
110
+ {
111
+ "updates": 950,
112
+ "consumed_samples": 121600,
113
+ "train_loss": 0.0650544323772192,
114
+ "validation_loss": 0.08221764412883203
115
+ },
116
+ {
117
+ "updates": 1000,
118
+ "consumed_samples": 128000,
119
+ "train_loss": 0.06138796729966998,
120
+ "validation_loss": 0.09766658151056617
121
+ },
122
+ {
123
+ "updates": 1050,
124
+ "consumed_samples": 134400,
125
+ "train_loss": 0.06051177423447371,
126
+ "validation_loss": 0.09287907368707238
127
+ },
128
+ {
129
+ "updates": 1100,
130
+ "consumed_samples": 140800,
131
+ "train_loss": 0.06050728471949696,
132
+ "validation_loss": 0.09408151025127154
133
+ },
134
+ {
135
+ "updates": 1150,
136
+ "consumed_samples": 147200,
137
+ "train_loss": 0.05924925757572055,
138
+ "validation_loss": 0.0946587849612115
139
+ },
140
+ {
141
+ "updates": 1200,
142
+ "consumed_samples": 153600,
143
+ "train_loss": 0.05794040573760867,
144
+ "validation_loss": 0.086535127764364
145
+ },
146
+ {
147
+ "updates": 1250,
148
+ "consumed_samples": 160000,
149
+ "train_loss": 0.056980984136462215,
150
+ "validation_loss": 0.09586095524718985
151
+ },
152
+ {
153
+ "updates": 1300,
154
+ "consumed_samples": 166400,
155
+ "train_loss": 0.05556875738315284,
156
+ "validation_loss": 0.09033180055121193
157
+ },
158
+ {
159
+ "updates": 1350,
160
+ "consumed_samples": 172800,
161
+ "train_loss": 0.0548527988512069,
162
+ "validation_loss": 0.08277507391176187
163
+ },
164
+ {
165
+ "updates": 1400,
166
+ "consumed_samples": 179200,
167
+ "train_loss": 0.05149021510966122,
168
+ "validation_loss": 0.09096652059088228
169
+ },
170
+ {
171
+ "updates": 1450,
172
+ "consumed_samples": 185600,
173
+ "train_loss": 0.051593654453754426,
174
+ "validation_loss": 0.08252263164467877
175
+ },
176
+ {
177
+ "updates": 1500,
178
+ "consumed_samples": 192000,
179
+ "train_loss": 0.05423289206810296,
180
+ "validation_loss": 0.08502114539805916
181
+ },
182
+ {
183
+ "updates": 1550,
184
+ "consumed_samples": 198400,
185
+ "train_loss": 0.052225465895608066,
186
+ "validation_loss": 0.08387484660488553
187
+ },
188
+ {
189
+ "updates": 1600,
190
+ "consumed_samples": 204800,
191
+ "train_loss": 0.05092524323612452,
192
+ "validation_loss": 0.08130560005520238
193
+ },
194
+ {
195
+ "updates": 1650,
196
+ "consumed_samples": 211200,
197
+ "train_loss": 0.04985399120487273,
198
+ "validation_loss": 0.08638887755660107
199
+ },
200
+ {
201
+ "updates": 1700,
202
+ "consumed_samples": 217600,
203
+ "train_loss": 0.049970695096999404,
204
+ "validation_loss": 0.0850406155550445
205
+ },
206
+ {
207
+ "updates": 1750,
208
+ "consumed_samples": 224000,
209
+ "train_loss": 0.04911648639477789,
210
+ "validation_loss": 0.08992611703433795
211
+ },
212
+ {
213
+ "updates": 1800,
214
+ "consumed_samples": 230400,
215
+ "train_loss": 0.04910631316713989,
216
+ "validation_loss": 0.09131536486529512
217
+ },
218
+ {
219
+ "updates": 1850,
220
+ "consumed_samples": 236800,
221
+ "train_loss": 0.045382359167560934,
222
+ "validation_loss": 0.10002207677462138
223
+ },
224
+ {
225
+ "updates": 1900,
226
+ "consumed_samples": 243200,
227
+ "train_loss": 0.0477118232473731,
228
+ "validation_loss": 0.10784883338783402
229
+ },
230
+ {
231
+ "updates": 1950,
232
+ "consumed_samples": 249600,
233
+ "train_loss": 0.046187796629965305,
234
+ "validation_loss": 0.09228424400498625
235
+ },
236
+ {
237
+ "updates": 2000,
238
+ "consumed_samples": 256000,
239
+ "train_loss": 0.04625743763521314,
240
+ "validation_loss": 0.0994702431744372
241
+ },
242
+ {
243
+ "updates": 2050,
244
+ "consumed_samples": 262400,
245
+ "train_loss": 0.04539411242119968,
246
+ "validation_loss": 0.10146190574960201
247
+ },
248
+ {
249
+ "updates": 2100,
250
+ "consumed_samples": 268800,
251
+ "train_loss": 0.044528810493648055,
252
+ "validation_loss": 0.10060195273399586
253
+ },
254
+ {
255
+ "updates": 2150,
256
+ "consumed_samples": 275200,
257
+ "train_loss": 0.04520396947860718,
258
+ "validation_loss": 0.11389675566897495
259
+ },
260
+ {
261
+ "updates": 2200,
262
+ "consumed_samples": 281600,
263
+ "train_loss": 0.04362621389329433,
264
+ "validation_loss": 0.09954333811765537
265
+ },
266
+ {
267
+ "updates": 2250,
268
+ "consumed_samples": 288000,
269
+ "train_loss": 0.042039215536788105,
270
+ "validation_loss": 0.09996076331299264
271
+ },
272
+ {
273
+ "updates": 2300,
274
+ "consumed_samples": 294400,
275
+ "train_loss": 0.04445314461365342,
276
+ "validation_loss": 0.09273883861897048
277
+ },
278
+ {
279
+ "updates": 2350,
280
+ "consumed_samples": 300800,
281
+ "train_loss": 0.044532884266227486,
282
+ "validation_loss": 0.10043427559867268
283
+ },
284
+ {
285
+ "updates": 2400,
286
+ "consumed_samples": 307200,
287
+ "train_loss": 0.043108444632962345,
288
+ "validation_loss": 0.09587047756940592
289
+ },
290
+ {
291
+ "updates": 2450,
292
+ "consumed_samples": 313600,
293
+ "train_loss": 0.04166162983514368,
294
+ "validation_loss": 0.09341833008511458
295
+ },
296
+ {
297
+ "updates": 2500,
298
+ "consumed_samples": 320000,
299
+ "train_loss": 0.04234295138157904,
300
+ "validation_loss": 0.09380867058280273
301
+ }
302
+ ]