Hugging Face's logo

cagataydev
/
gr00t-wave

Robotics
Transformers
TensorBoard
Safetensors
English
gr00t_n1_5
foundation-model
gr00t
dual-camera
robot-learning
manipulation
embodied-ai
Model card Files
xet
Metrics Community
cagataydev commited on
Commit
598d68e
·
verified ·
1 Parent(s): e5f4a43

Upload folder using huggingface_hub

Browse files
Files changed (26) hide show
  1. checkpoint-150000/config.json +64 -0
  2. checkpoint-150000/experiment_cfg/metadata.json +195 -0
  3. checkpoint-150000/model-00001-of-00002.safetensors +3 -0
  4. checkpoint-150000/model-00002-of-00002.safetensors +3 -0
  5. checkpoint-150000/model.safetensors.index.json +0 -0
  6. checkpoint-150000/optimizer.pt +3 -0
  7. checkpoint-150000/rng_state.pth +3 -0
  8. checkpoint-150000/scheduler.pt +3 -0
  9. checkpoint-150000/trainer_state.json +0 -0
  10. checkpoint-300000/config.json +64 -0
  11. checkpoint-300000/experiment_cfg/metadata.json +195 -0
  12. checkpoint-300000/model-00001-of-00002.safetensors +3 -0
  13. checkpoint-300000/model-00002-of-00002.safetensors +3 -0
  14. checkpoint-300000/model.safetensors.index.json +0 -0
  15. checkpoint-300000/optimizer.pt +3 -0
  16. checkpoint-300000/rng_state.pth +3 -0
  17. checkpoint-300000/scheduler.pt +3 -0
  18. checkpoint-300000/trainer_state.json +0 -0
  19. config.json +64 -0
  20. experiment_cfg/metadata.json +195 -0
  21. model-00001-of-00002.safetensors +3 -0
  22. model-00002-of-00002.safetensors +3 -0
  23. model.safetensors.index.json +0 -0
  24. runs/Sep29_04-23-03_ip-172-31-3-77/events.out.tfevents.1759119788.ip-172-31-3-77.695709.0 +3 -0
  25. trainer_state.json +0 -0
  26. training_args.bin +3 -0
checkpoint-150000/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-150000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 66.4754638671875,
8
+ 54.32937240600586,
9
+ 99.81834411621094,
10
+ 99.04306030273438,
11
+ 1.431918740272522
12
+ ],
13
+ "min": [
14
+ -57.29764175415039,
15
+ -99.06621551513672,
16
+ -96.91189575195312,
17
+ -71.37886047363281,
18
+ -77.45378875732422
19
+ ],
20
+ "mean": [
21
+ -0.8736613392829895,
22
+ -35.767276763916016,
23
+ 24.052417755126953,
24
+ 34.033084869384766,
25
+ -52.904090881347656
26
+ ],
27
+ "std": [
28
+ 15.214531898498535,
29
+ 36.241241455078125,
30
+ 48.20182418823242,
31
+ 32.278812408447266,
32
+ 7.842017650604248
33
+ ],
34
+ "q01": [
35
+ -40.98151779174805,
36
+ -98.8115463256836,
37
+ -84.10536193847656,
38
+ -49.45628356933594,
39
+ -68.57693481445312
40
+ ],
41
+ "q99": [
42
+ 40.85404586791992,
43
+ 34.25466949462894,
44
+ 99.6366958618164,
45
+ 90.91431152343762,
46
+ -29.44337371826167
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 7.900067329406738
52
+ ],
53
+ "min": [
54
+ 1.4179608821868896
55
+ ],
56
+ "mean": [
57
+ 2.2223942279815674
58
+ ],
59
+ "std": [
60
+ 0.8177542686462402
61
+ ],
62
+ "q01": [
63
+ 1.4179608821868896
64
+ ],
65
+ "q99": [
66
+ 3.4436192512512207
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 69.63151550292969,
74
+ 53.63175582885742,
75
+ 99.81908416748047,
76
+ 100.0,
77
+ -6.163849353790283
78
+ ],
79
+ "min": [
80
+ -61.245235443115234,
81
+ -100.0,
82
+ -100.0,
83
+ -76.24620819091797,
84
+ -80.18205261230469
85
+ ],
86
+ "mean": [
87
+ -0.7885429859161377,
88
+ -35.86227798461914,
89
+ 21.236610412597656,
90
+ 33.204769134521484,
91
+ -52.84980392456055
92
+ ],
93
+ "std": [
94
+ 15.83127498626709,
95
+ 35.13792037963867,
96
+ 48.32182693481445,
97
+ 33.81686782836914,
98
+ 8.233314514160156
99
+ ],
100
+ "q01": [
101
+ -42.94790267944336,
102
+ -99.15540313720703,
103
+ -87.0646743774414,
104
+ -53.09926223754883,
105
+ -69.91001403808593
106
+ ],
107
+ "q99": [
108
+ 43.07496643066406,
109
+ 32.56081130981449,
110
+ 99.72863006591797,
111
+ 92.37104797363281,
112
+ -27.567230377197244
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 8.129439353942871
118
+ ],
119
+ "min": [
120
+ 1.499605417251587
121
+ ],
122
+ "mean": [
123
+ 2.252654552459717
124
+ ],
125
+ "std": [
126
+ 0.827427864074707
127
+ ],
128
+ "q01": [
129
+ 1.499605417251587
130
+ ],
131
+ "q99": [
132
+ 3.709550142288208
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-150000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc9070e30d207d0aabb8ed7fe831892f105317965556fa512f28a990f03ed28b
3
+ size 4999367032
checkpoint-150000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dda87e0ea199fe69cdc506562b8abffc5ad610a5cf9b88eb78f4f43c9406e6a6
3
+ size 2586705312
checkpoint-150000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-150000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e7a53e14b9a2ae6991b604435ace9dc3d6d37e7ac4069f0485f0331858a247e
3
+ size 8550720062
checkpoint-150000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fe10739fb2f268e03ee32db05816a46a101d9c49ffe7109942afea2968d89f8
3
+ size 14244
checkpoint-150000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d1a646f8b9d000782dffa1e092b38bbd22100c8d80a1aa74ec7edb89194d0f5
3
+ size 1064
checkpoint-150000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-300000/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-300000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 66.4754638671875,
8
+ 54.32937240600586,
9
+ 99.81834411621094,
10
+ 99.04306030273438,
11
+ 1.431918740272522
12
+ ],
13
+ "min": [
14
+ -57.29764175415039,
15
+ -99.06621551513672,
16
+ -96.91189575195312,
17
+ -71.37886047363281,
18
+ -77.45378875732422
19
+ ],
20
+ "mean": [
21
+ -0.8736613392829895,
22
+ -35.767276763916016,
23
+ 24.052417755126953,
24
+ 34.033084869384766,
25
+ -52.904090881347656
26
+ ],
27
+ "std": [
28
+ 15.214531898498535,
29
+ 36.241241455078125,
30
+ 48.20182418823242,
31
+ 32.278812408447266,
32
+ 7.842017650604248
33
+ ],
34
+ "q01": [
35
+ -40.98151779174805,
36
+ -98.8115463256836,
37
+ -84.10536193847656,
38
+ -49.45628356933594,
39
+ -68.57693481445312
40
+ ],
41
+ "q99": [
42
+ 40.85404586791992,
43
+ 34.25466949462894,
44
+ 99.6366958618164,
45
+ 90.91431152343762,
46
+ -29.44337371826167
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 7.900067329406738
52
+ ],
53
+ "min": [
54
+ 1.4179608821868896
55
+ ],
56
+ "mean": [
57
+ 2.2223942279815674
58
+ ],
59
+ "std": [
60
+ 0.8177542686462402
61
+ ],
62
+ "q01": [
63
+ 1.4179608821868896
64
+ ],
65
+ "q99": [
66
+ 3.4436192512512207
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 69.63151550292969,
74
+ 53.63175582885742,
75
+ 99.81908416748047,
76
+ 100.0,
77
+ -6.163849353790283
78
+ ],
79
+ "min": [
80
+ -61.245235443115234,
81
+ -100.0,
82
+ -100.0,
83
+ -76.24620819091797,
84
+ -80.18205261230469
85
+ ],
86
+ "mean": [
87
+ -0.7885429859161377,
88
+ -35.86227798461914,
89
+ 21.236610412597656,
90
+ 33.204769134521484,
91
+ -52.84980392456055
92
+ ],
93
+ "std": [
94
+ 15.83127498626709,
95
+ 35.13792037963867,
96
+ 48.32182693481445,
97
+ 33.81686782836914,
98
+ 8.233314514160156
99
+ ],
100
+ "q01": [
101
+ -42.94790267944336,
102
+ -99.15540313720703,
103
+ -87.0646743774414,
104
+ -53.09926223754883,
105
+ -69.91001403808593
106
+ ],
107
+ "q99": [
108
+ 43.07496643066406,
109
+ 32.56081130981449,
110
+ 99.72863006591797,
111
+ 92.37104797363281,
112
+ -27.567230377197244
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 8.129439353942871
118
+ ],
119
+ "min": [
120
+ 1.499605417251587
121
+ ],
122
+ "mean": [
123
+ 2.252654552459717
124
+ ],
125
+ "std": [
126
+ 0.827427864074707
127
+ ],
128
+ "q01": [
129
+ 1.499605417251587
130
+ ],
131
+ "q99": [
132
+ 3.709550142288208
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-300000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce1d91c65a94d1d2d73348a2c72d34d1d2fe0b53c6332a5c29792f4ee494215a
3
+ size 4999367032
checkpoint-300000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d93630402a71d67cbc79f6205e41cd57d6f8e25ed2061386300193083e8ed56
3
+ size 2586705312
checkpoint-300000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-300000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54ce326ac14ebf25ba89ed75809d5762b52f6bc81f959048f9965fab7b400299
3
+ size 8550720062
checkpoint-300000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c389964b1f66a85a01287f0c6d0d12b508a71e35452f01a4b891e43fb72e1f87
3
+ size 14244
checkpoint-300000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f221bc96f6ca122af3955544cbd22d3fbf7cbc82347cb843127139df2c7febd
3
+ size 1064
checkpoint-300000/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
+ 66.4754638671875,
8
+ 54.32937240600586,
9
+ 99.81834411621094,
10
+ 99.04306030273438,
11
+ 1.431918740272522
12
+ ],
13
+ "min": [
14
+ -57.29764175415039,
15
+ -99.06621551513672,
16
+ -96.91189575195312,
17
+ -71.37886047363281,
18
+ -77.45378875732422
19
+ ],
20
+ "mean": [
21
+ -0.8736613392829895,
22
+ -35.767276763916016,
23
+ 24.052417755126953,
24
+ 34.033084869384766,
25
+ -52.904090881347656
26
+ ],
27
+ "std": [
28
+ 15.214531898498535,
29
+ 36.241241455078125,
30
+ 48.20182418823242,
31
+ 32.278812408447266,
32
+ 7.842017650604248
33
+ ],
34
+ "q01": [
35
+ -40.98151779174805,
36
+ -98.8115463256836,
37
+ -84.10536193847656,
38
+ -49.45628356933594,
39
+ -68.57693481445312
40
+ ],
41
+ "q99": [
42
+ 40.85404586791992,
43
+ 34.25466949462894,
44
+ 99.6366958618164,
45
+ 90.91431152343762,
46
+ -29.44337371826167
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 7.900067329406738
52
+ ],
53
+ "min": [
54
+ 1.4179608821868896
55
+ ],
56
+ "mean": [
57
+ 2.2223942279815674
58
+ ],
59
+ "std": [
60
+ 0.8177542686462402
61
+ ],
62
+ "q01": [
63
+ 1.4179608821868896
64
+ ],
65
+ "q99": [
66
+ 3.4436192512512207
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 69.63151550292969,
74
+ 53.63175582885742,
75
+ 99.81908416748047,
76
+ 100.0,
77
+ -6.163849353790283
78
+ ],
79
+ "min": [
80
+ -61.245235443115234,
81
+ -100.0,
82
+ -100.0,
83
+ -76.24620819091797,
84
+ -80.18205261230469
85
+ ],
86
+ "mean": [
87
+ -0.7885429859161377,
88
+ -35.86227798461914,
89
+ 21.236610412597656,
90
+ 33.204769134521484,
91
+ -52.84980392456055
92
+ ],
93
+ "std": [
94
+ 15.83127498626709,
95
+ 35.13792037963867,
96
+ 48.32182693481445,
97
+ 33.81686782836914,
98
+ 8.233314514160156
99
+ ],
100
+ "q01": [
101
+ -42.94790267944336,
102
+ -99.15540313720703,
103
+ -87.0646743774414,
104
+ -53.09926223754883,
105
+ -69.91001403808593
106
+ ],
107
+ "q99": [
108
+ 43.07496643066406,
109
+ 32.56081130981449,
110
+ 99.72863006591797,
111
+ 92.37104797363281,
112
+ -27.567230377197244
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 8.129439353942871
118
+ ],
119
+ "min": [
120
+ 1.499605417251587
121
+ ],
122
+ "mean": [
123
+ 2.252654552459717
124
+ ],
125
+ "std": [
126
+ 0.827427864074707
127
+ ],
128
+ "q01": [
129
+ 1.499605417251587
130
+ ],
131
+ "q99": [
132
+ 3.709550142288208
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:ce1d91c65a94d1d2d73348a2c72d34d1d2fe0b53c6332a5c29792f4ee494215a
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d93630402a71d67cbc79f6205e41cd57d6f8e25ed2061386300193083e8ed56
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
runs/Sep29_04-23-03_ip-172-31-3-77/events.out.tfevents.1759119788.ip-172-31-3-77.695709.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87e1c17956245ca726c2c219e0d4f62c564bbf3297b57029643f16bf373570c0
3
+ size 6449912
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:051235663afdaf4bf860b0f0e27b5f9b5f81fd8b9a086a15b6623c23c39f1abc
3
+ size 5368