Hugging Face's logo

shadowHokage
/
groot_trained

Safetensors
gr00t_n1_5
Model card Files
xet
Community
shadowHokage commited on
Commit
354094b
·
verified ·
1 Parent(s): f7f34a3

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. checkpoint-10000/config.json +64 -0
  2. checkpoint-10000/experiment_cfg/metadata.json +195 -0
  3. checkpoint-10000/model-00001-of-00002.safetensors +3 -0
  4. checkpoint-10000/model-00002-of-00002.safetensors +3 -0
  5. checkpoint-10000/model.safetensors.index.json +0 -0
  6. checkpoint-10000/optimizer.pt +3 -0
  7. checkpoint-10000/rng_state.pth +3 -0
  8. checkpoint-10000/scheduler.pt +3 -0
  9. checkpoint-10000/trainer_state.json +0 -0
  10. checkpoint-6000/config.json +64 -0
  11. checkpoint-6000/experiment_cfg/metadata.json +195 -0
  12. checkpoint-6000/model-00001-of-00002.safetensors +3 -0
  13. checkpoint-6000/model-00002-of-00002.safetensors +3 -0
  14. checkpoint-6000/model.safetensors.index.json +0 -0
  15. checkpoint-6000/optimizer.pt +3 -0
  16. checkpoint-6000/rng_state.pth +3 -0
  17. checkpoint-6000/scheduler.pt +3 -0
  18. checkpoint-6000/trainer_state.json +0 -0
  19. checkpoint-7000/config.json +64 -0
  20. checkpoint-7000/experiment_cfg/metadata.json +195 -0
  21. checkpoint-7000/model-00001-of-00002.safetensors +3 -0
  22. checkpoint-7000/model-00002-of-00002.safetensors +3 -0
  23. checkpoint-7000/model.safetensors.index.json +0 -0
  24. checkpoint-7000/optimizer.pt +3 -0
  25. checkpoint-7000/rng_state.pth +3 -0
  26. checkpoint-7000/scheduler.pt +3 -0
  27. checkpoint-7000/trainer_state.json +0 -0
  28. checkpoint-8000/config.json +64 -0
  29. checkpoint-8000/experiment_cfg/metadata.json +195 -0
  30. checkpoint-8000/model-00001-of-00002.safetensors +3 -0
  31. checkpoint-8000/model-00002-of-00002.safetensors +3 -0
  32. checkpoint-8000/model.safetensors.index.json +0 -0
  33. checkpoint-8000/optimizer.pt +3 -0
  34. checkpoint-8000/rng_state.pth +3 -0
  35. checkpoint-8000/scheduler.pt +3 -0
  36. checkpoint-8000/trainer_state.json +0 -0
  37. checkpoint-9000/config.json +64 -0
  38. checkpoint-9000/experiment_cfg/metadata.json +195 -0
  39. checkpoint-9000/model-00001-of-00002.safetensors +3 -0
  40. checkpoint-9000/model-00002-of-00002.safetensors +3 -0
  41. checkpoint-9000/model.safetensors.index.json +0 -0
  42. checkpoint-9000/optimizer.pt +3 -0
  43. checkpoint-9000/rng_state.pth +3 -0
  44. checkpoint-9000/scheduler.pt +3 -0
  45. checkpoint-9000/trainer_state.json +0 -0
  46. config.json +64 -0
  47. experiment_cfg/metadata.json +195 -0
  48. model-00001-of-00002.safetensors +3 -0
  49. model-00002-of-00002.safetensors +3 -0
  50. model.safetensors.index.json +0 -0
checkpoint-10000/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
checkpoint-10000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
checkpoint-10000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99716f927323c807b43988fe0f87dca790aff2ace0700497032f3a151a867557
3
+ size 4999367032
checkpoint-10000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c5dfc1c79afe9228d128c226a890b72ee7e8f1091405abf515f712414618377
3
+ size 2586705312
checkpoint-10000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86305ca4efc9f0fe12078a6b4452f3bc9e357614c347e8943a6d298f1a6bee30
3
+ size 8550720062
checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d012fbd6e5db9902d55ec0cd7442fb83c2a8efd108d0510045f3e27b1a60d6c3
3
+ size 14244
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07356cb668b8efc440ddb2f7f81f7da3152060e81d0229922fa3b30df3313f82
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-6000/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
checkpoint-6000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
checkpoint-6000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3724ecd5d8816044c5ee862f282f7b00084aef3cf70f664952ddc36b47e1b77
3
+ size 4999367032
checkpoint-6000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eebc24afb94fef198fe47dd39d0fb1cc08882e5cfdec2311cd986e3b00f7722
3
+ size 2586705312
checkpoint-6000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-6000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc1aeec8d80f0ac60a5768f32d48c5fea975993ed5a26306752e9b7316a7d9f5
3
+ size 8550720062
checkpoint-6000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc9c3bd2097ef84b426f1129f9d3d2f4b9e1abc099cc58076ba6224226ee96d1
3
+ size 14244
checkpoint-6000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d6c0bf646ac78d188a2258804081d2e369772e8091e5395d9073771ec310aca
3
+ size 1064
checkpoint-6000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-7000/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
checkpoint-7000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
checkpoint-7000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:463cbb805939758ace11f0f5f2d3d00b5f8068336f1c405f9e7e44a0cf7350e4
3
+ size 4999367032
checkpoint-7000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:289d309bf7b29642f8d5bb45cde1c51f095b46aaf8725fe0facab5c44d983d6b
3
+ size 2586705312
checkpoint-7000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-7000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d11f5c865f5e3db98b60d1a74a2c15ad1f736af2ffe34cd8022200797758b46c
3
+ size 8550720062
checkpoint-7000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d508c656d43c155e609fcf6c730a73bac873e5f4c673ad68eb8f43bc1b8c96b0
3
+ size 14244
checkpoint-7000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45049460c0fa55fae34b1884c511d0c5cb5241d0d636fad92c82b0e0a657580e
3
+ size 1064
checkpoint-7000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-8000/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
checkpoint-8000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
checkpoint-8000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92c4ecc1aa440abc3c0b4145a6751930397bde9893e4fcb27178788e29af67de
3
+ size 4999367032
checkpoint-8000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9487e54ecba7427b0559091b4ff3ee925b6ffcae4c0ef982eb123a746888a0df
3
+ size 2586705312
checkpoint-8000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-8000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:290b267195db854095c701538fbd060ffaad978aaef474d873c68b54b994b369
3
+ size 8550720062
checkpoint-8000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6dc4125a87787ba817aa92c23dd796ad0b85199bb61d4e8b837ed424a54e7bcb
3
+ size 14244
checkpoint-8000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00739e77892b3b90e07da5aa63424b8c4adff832abcfc91f97c4f8838af650db
3
+ size 1064
checkpoint-8000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
checkpoint-9000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
checkpoint-9000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:066e5d304c5dd7dcf408934677652a4dbcec6f01240e1dda187a70ce6bd60aba
3
+ size 4999367032
checkpoint-9000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32e7fae21327c48363bf115485853fbd1bcbfdcb94ce5ce998e259d9ac36062c
3
+ size 2586705312
checkpoint-9000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:def23c535b8e13f259dbd1c6b60ab8506fab8837adc1f5e81789d7d8116aa331
3
+ size 8550720062
checkpoint-9000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5905008cd02e8a06093b9d9c465c32f9543c3ea5501f207b79aec3858eef6701
3
+ size 14244
checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c26823aa5685497db4fe82a648513d12cb6c26d9be110887b197c10fe3a44ac
3
+ size 1064
checkpoint-9000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "torch_dtype": "bfloat16",
63
+ "transformers_version": "4.51.3"
64
+ }
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 52.71546936035156,
8
+ 65.318359375,
9
+ 100.1607666015625,
10
+ 100.02542114257812,
11
+ 50.691986083984375
12
+ ],
13
+ "min": [
14
+ -38.4887809753418,
15
+ -99.55461120605469,
16
+ -96.33113098144531,
17
+ 16.11701202392578,
18
+ -14.437675476074219
19
+ ],
20
+ "mean": [
21
+ 8.840156555175781,
22
+ 4.428213119506836,
23
+ 3.7019574642181396,
24
+ 75.59637451171875,
25
+ 18.457332611083984
26
+ ],
27
+ "std": [
28
+ 19.552515029907227,
29
+ 39.316009521484375,
30
+ 41.37403106689453,
31
+ 13.751623153686523,
32
+ 9.614367485046387
33
+ ],
34
+ "q01": [
35
+ -24.408675842285156,
36
+ -98.82302276611328,
37
+ -69.73289306640625,
38
+ 39.07359924316406,
39
+ -4.678773803710937
40
+ ],
41
+ "q99": [
42
+ 47.0116217041016,
43
+ 60.00616455078126,
44
+ 93.92067687988282,
45
+ 99.91351684570313,
46
+ 39.71997375488284
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 88.45756530761719
52
+ ],
53
+ "min": [
54
+ 2.1588988304138184
55
+ ],
56
+ "mean": [
57
+ 33.337127685546875
58
+ ],
59
+ "std": [
60
+ 11.978549003601074
61
+ ],
62
+ "q01": [
63
+ 2.5056805324554445
64
+ ],
65
+ "q99": [
66
+ 58.33354202270508
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 52.056304931640625,
74
+ 63.758392333984375,
75
+ 99.45799255371094,
76
+ 100.0,
77
+ 51.0172119140625
78
+ ],
79
+ "min": [
80
+ -38.495731353759766,
81
+ -99.99999237060547,
82
+ -99.1869888305664,
83
+ 21.52564239501953,
84
+ -14.449661254882812
85
+ ],
86
+ "mean": [
87
+ 8.859687805175781,
88
+ 2.965143918991089,
89
+ 2.249222993850708,
90
+ 76.57550048828125,
91
+ 18.281843185424805
92
+ ],
93
+ "std": [
94
+ 19.56995391845703,
95
+ 39.62527084350586,
96
+ 41.96939468383789,
97
+ 13.673238754272461,
98
+ 9.612025260925293
99
+ ],
100
+ "q01": [
101
+ -24.194137573242188,
102
+ -99.16107177734375,
103
+ -70.280029296875,
104
+ 40.99079895019531,
105
+ -4.955657958984375
106
+ ],
107
+ "q99": [
108
+ 46.72100830078125,
109
+ 58.47315979003906,
110
+ 92.86360168457031,
111
+ 99.82463073730469,
112
+ 39.8017578125
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 91.3986587524414
118
+ ],
119
+ "min": [
120
+ 0.5983543395996094
121
+ ],
122
+ "mean": [
123
+ 32.5671501159668
124
+ ],
125
+ "std": [
126
+ 12.373101234436035
127
+ ],
128
+ "q01": [
129
+ 2.393418788909912
130
+ ],
131
+ "q99": [
132
+ 58.33956527709961
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "front": {
140
+ "resolution": [
141
+ 640,
142
+ 480
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "wrist": {
148
+ "resolution": [
149
+ 640,
150
+ 480
151
+ ],
152
+ "channels": 3,
153
+ "fps": 30.0
154
+ }
155
+ },
156
+ "state": {
157
+ "single_arm": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 5
162
+ ],
163
+ "continuous": true
164
+ },
165
+ "gripper": {
166
+ "absolute": true,
167
+ "rotation_type": null,
168
+ "shape": [
169
+ 1
170
+ ],
171
+ "continuous": true
172
+ }
173
+ },
174
+ "action": {
175
+ "single_arm": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 5
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 1
188
+ ],
189
+ "continuous": true
190
+ }
191
+ }
192
+ },
193
+ "embodiment_tag": "new_embodiment"
194
+ }
195
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99716f927323c807b43988fe0f87dca790aff2ace0700497032f3a151a867557
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c5dfc1c79afe9228d128c226a890b72ee7e8f1091405abf515f712414618377
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff