Hugging Face's logo

tfaith
/
sclab-so101-1

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
tfaith commited on
Commit
43b4a76
·
verified ·
1 Parent(s): f77ff66

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
+ 19.16030502319336,
8
+ 79.8241958618164,
9
+ 99.7272720336914,
10
+ 99.48936462402344,
11
+ 50.525760650634766
12
+ ],
13
+ "min": [
14
+ -56.64122009277344,
15
+ -99.16282653808594,
16
+ -97.09091186523438,
17
+ -84.17021179199219,
18
+ -92.95478820800781
19
+ ],
20
+ "mean": [
21
+ -13.593049049377441,
22
+ -37.8803825378418,
23
+ 49.65174865722656,
24
+ 41.38975524902344,
25
+ -35.52872085571289
26
+ ],
27
+ "std": [
28
+ 18.09169578552246,
29
+ 52.506229400634766,
30
+ 47.15378189086914,
31
+ 38.462406158447266,
32
+ 40.10171127319336
33
+ ],
34
+ "q01": [
35
+ -53.435115814208984,
36
+ -98.91168212890625,
37
+ -77.36363983154297,
38
+ -64.0851058959961,
39
+ -85.43638610839844
40
+ ],
41
+ "q99": [
42
+ 16.64122200012207,
43
+ 61.21808338165283,
44
+ 99.63636016845703,
45
+ 95.06382751464844,
46
+ 50.15773010253906
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 47.49163818359375
52
+ ],
53
+ "min": [
54
+ 1.471571922302246
55
+ ],
56
+ "mean": [
57
+ 13.221816062927246
58
+ ],
59
+ "std": [
60
+ 12.446600914001465
61
+ ],
62
+ "q01": [
63
+ 1.471571922302246
64
+ ],
65
+ "q99": [
66
+ 45.0167236328125
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 19.560440063476562,
74
+ 79.14059448242188,
75
+ 99.81965637207031,
76
+ 100.0,
77
+ 50.64321517944336
78
+ ],
79
+ "min": [
80
+ -57.28937911987305,
81
+ -100.0,
82
+ -99.72948455810547,
83
+ -85.26405334472656,
84
+ -93.961669921875
85
+ ],
86
+ "mean": [
87
+ -13.568873405456543,
88
+ -38.900142669677734,
89
+ 48.5997428894043,
90
+ 41.279808044433594,
91
+ -35.4924430847168
92
+ ],
93
+ "std": [
94
+ 18.174137115478516,
95
+ 52.097694396972656,
96
+ 47.721221923828125,
97
+ 38.7339973449707,
98
+ 40.1052360534668
99
+ ],
100
+ "q01": [
101
+ -53.64468860626221,
102
+ -99.58280944824219,
103
+ -79.17042541503906,
104
+ -65.33219909667969,
105
+ -85.61302185058594
106
+ ],
107
+ "q99": [
108
+ 16.8681321144104,
109
+ 59.13642120361328,
110
+ 99.63931274414062,
111
+ 95.3151626586914,
112
+ 50.2231559753418
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 48.13916015625
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 11.667485237121582
124
+ ],
125
+ "std": [
126
+ 13.583215713500977
127
+ ],
128
+ "q01": [
129
+ 0.8090614676475525
130
+ ],
131
+ "q99": [
132
+ 44.90291213989258
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "overhead": {
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:d98754247d87ee0844ad9d2aa22099d56270f52382a484cc0a5b90a258f7e972
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ee694c1589647a0608ec0044f7d2ab25a459edda188a018b80eefe88f16fb3b
3
+ size 2586705312
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:ad177cb09964a600c8f2d145aef6da2ddb70cf559171afaf2a7e03eec3c49e98
3
+ size 8550720062
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afb454c73a8367fcfd910d27f077ea30ffbe48b21010a303e0236039077e6735
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07356cb668b8efc440ddb2f7f81f7da3152060e81d0229922fa3b30df3313f82
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff