Hugging Face's logo

nvidia
/
SO_ARM_Starter_Gr00t

Safetensors
gr00t_n1_5
Model card Files
xet
Community
2
mingxinz commited on
Commit
c10f391
·
verified ·
1 Parent(s): 4248b05

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +195 -0
  3. model-00001-of-00002.safetensors +3 -0
  4. model-00002-of-00002.safetensors +3 -0
  5. model.safetensors.index.json +0 -0
  6. optimizer.pt +3 -0
  7. rng_state.pth +3 -0
  8. scheduler.pt +3 -0
  9. trainer_state.json +0 -0
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
+ 35.86289978027344,
8
+ 50.3662223815918,
9
+ 102.207763671875,
10
+ 90.01766967773438,
11
+ -31.214004516601562
12
+ ],
13
+ "min": [
14
+ -29.695632934570312,
15
+ -100.08529663085938,
16
+ -45.66314697265625,
17
+ -40.74222946166992,
18
+ -86.94236755371094
19
+ ],
20
+ "mean": [
21
+ 2.7533340252980514,
22
+ -44.83029396651267,
23
+ 60.342883870794346,
24
+ 40.597277393208074,
25
+ -50.334554287325034
26
+ ],
27
+ "std": [
28
+ 17.734981201267647,
29
+ 49.17619133821294,
30
+ 38.56363047660454,
31
+ 24.53213903612298,
32
+ 8.054818688839257
33
+ ],
34
+ "q01": [
35
+ -23.21019086928238,
36
+ -99.89977273843078,
37
+ -11.41253884822063,
38
+ -6.254064539984565,
39
+ -70.48279780022536
40
+ ],
41
+ "q99": [
42
+ 31.90897636539444,
43
+ 27.867335486293534,
44
+ 100.98642284975836,
45
+ 81.64018151404724,
46
+ -36.26702092112002
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 35.36993408203125
52
+ ],
53
+ "min": [
54
+ 0.8484606742858887
55
+ ],
56
+ "mean": [
57
+ 5.293151230143854
58
+ ],
59
+ "std": [
60
+ 6.097173548270028
61
+ ],
62
+ "q01": [
63
+ 1.0422283546994597
64
+ ],
65
+ "q99": [
66
+ 24.367772452631534
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 41.80928039550781,
74
+ 49.574562072753906,
75
+ 100.0,
76
+ 90.16828918457031,
77
+ -29.435836791992188
78
+ ],
79
+ "min": [
80
+ -33.398529052734375,
81
+ -100.0,
82
+ -48.1590576171875,
83
+ -33.30376434326172,
84
+ -88.4244384765625
85
+ ],
86
+ "mean": [
87
+ 2.5380374790115976,
88
+ -46.0233579757557,
89
+ 59.28242407975492,
90
+ 40.79325100500621,
91
+ -50.390393499610205
92
+ ],
93
+ "std": [
94
+ 18.187261295431064,
95
+ 50.03983853983534,
96
+ 39.938257415119935,
97
+ 25.29606487293311,
98
+ 8.806160091188778
99
+ ],
100
+ "q01": [
101
+ -24.114472398492893,
102
+ -99.8795200787323,
103
+ -17.358844610605892,
104
+ -7.204820546773284,
105
+ -71.35633351223814
106
+ ],
107
+ "q99": [
108
+ 32.89371661209723,
109
+ 28.987150456925733,
110
+ 99.93052443326673,
111
+ 79.60891411129576,
112
+ -34.8488872774947
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 39.285709381103516
118
+ ],
119
+ "min": [
120
+ 8.669766202729079e-07
121
+ ],
122
+ "mean": [
123
+ 4.878072070944325
124
+ ],
125
+ "std": [
126
+ 6.868800906720681
127
+ ],
128
+ "q01": [
129
+ 0.7697634228634059
130
+ ],
131
+ "q99": [
132
+ 27.233317070528038
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "room": {
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 sha256:3a14b8aa7ac4b0e1746c2a525517799c442d6c14601ac9a8d4541a0794c3be18
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:258d026956cbe1912691bbcce61aac4e044ad0e111da3c452ccae44b79afbe04
3
+ size 2586508600
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6797eee91c406ace67fe8504c871cf4d4857d6dcc5b72edb93e209408dc3b3d7
3
+ size 8550325978
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24806fb3ff873a66d3f705a6c87625872d2d31a62185048c71e976e3ab453021
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe8c620e9a7808e10f2bc64fa81fe5b1efb6dcc2acb3596ed32f58119a5368c1
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff