Hugging Face's logo

nextbig
/
gr00t-3b-sl-4-sa

Safetensors
gr00t_n1_5
Model card Files
xet
Community
nextbig commited on
Commit
48f59e7
·
verified ·
1 Parent(s): f29bb58

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +215 -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. trainer_state.json +0 -0
  7. training_args.bin +3 -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": "4.52.1"
64
+ }
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,215 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "right_arm": {
6
+ "max": [
7
+ 0.9610829949378967,
8
+ 2.7031660079956055,
9
+ 0.009651671163737774,
10
+ 1.6130506992340088,
11
+ 1.225936770439148,
12
+ 1.810639500617981
13
+ ],
14
+ "min": [
15
+ -0.2552544176578522,
16
+ -0.008639379404485226,
17
+ -2.7631354331970215,
18
+ -1.7382606267929077,
19
+ -0.6361026763916016,
20
+ -2.0484580993652344
21
+ ],
22
+ "mean": [
23
+ 0.3781740963459015,
24
+ 1.5795903205871582,
25
+ -1.283855676651001,
26
+ 0.11046278476715088,
27
+ 0.6009297966957092,
28
+ -0.2991737723350525
29
+ ],
30
+ "std": [
31
+ 0.29604291915893555,
32
+ 0.8378310203552246,
33
+ 0.8045353889465332,
34
+ 0.26887059211730957,
35
+ 0.4417983889579773,
36
+ 0.4677432179450989
37
+ ],
38
+ "q01": [
39
+ -0.12683308124542236,
40
+ 0.001006182329729203,
41
+ -2.637436866760254,
42
+ -0.6975173443555832,
43
+ -0.2836762100458145,
44
+ -1.5315613746643066
45
+ ],
46
+ "q99": [
47
+ 0.9188634753227234,
48
+ 2.6161787509918213,
49
+ 0.006928957067430019,
50
+ 0.9580077290534961,
51
+ 1.2202469110488892,
52
+ 0.9755640059709526
53
+ ]
54
+ },
55
+ "right_gripper": {
56
+ "max": [
57
+ 0.09276922792196274
58
+ ],
59
+ "min": [
60
+ -0.0008076922968029976
61
+ ],
62
+ "mean": [
63
+ 0.03244227170944214
64
+ ],
65
+ "std": [
66
+ 0.025363579392433167
67
+ ],
68
+ "q01": [
69
+ -0.0005769230774603784
70
+ ],
71
+ "q99": [
72
+ 0.09265384823083878
73
+ ]
74
+ }
75
+ },
76
+ "action": {
77
+ "right_arm": {
78
+ "max": [
79
+ 0.9610829949378967,
80
+ 2.7031660079956055,
81
+ 0.009651671163737774,
82
+ 1.6130506992340088,
83
+ 1.225936770439148,
84
+ 1.810639500617981
85
+ ],
86
+ "min": [
87
+ -0.2552544176578522,
88
+ -0.008639379404485226,
89
+ -2.7631354331970215,
90
+ -1.7382606267929077,
91
+ -0.6361026763916016,
92
+ -2.0484580993652344
93
+ ],
94
+ "mean": [
95
+ 0.37816980481147766,
96
+ 1.579582691192627,
97
+ -1.2838547229766846,
98
+ 0.11046114563941956,
99
+ 0.6009286046028137,
100
+ -0.2991686165332794
101
+ ],
102
+ "std": [
103
+ 0.2960471510887146,
104
+ 0.83784419298172,
105
+ 0.8045370578765869,
106
+ 0.26887476444244385,
107
+ 0.4418012499809265,
108
+ 0.46775001287460327
109
+ ],
110
+ "q01": [
111
+ -0.12683308124542236,
112
+ 0.001029744278639555,
113
+ -2.637436866760254,
114
+ -0.6975173443555832,
115
+ -0.2836762100458145,
116
+ -1.5315613746643066
117
+ ],
118
+ "q99": [
119
+ 0.9188634753227234,
120
+ 2.6161787509918213,
121
+ 0.006928957067430019,
122
+ 0.9580077290534961,
123
+ 1.2202469110488892,
124
+ 0.9755640059709526
125
+ ]
126
+ },
127
+ "right_gripper": {
128
+ "max": [
129
+ 0.09276922792196274
130
+ ],
131
+ "min": [
132
+ -0.0008076922968029976
133
+ ],
134
+ "mean": [
135
+ 0.032442253082990646
136
+ ],
137
+ "std": [
138
+ 0.025363612920045853
139
+ ],
140
+ "q01": [
141
+ -0.0005769230774603784
142
+ ],
143
+ "q99": [
144
+ 0.09265384823083878
145
+ ]
146
+ }
147
+ }
148
+ },
149
+ "modalities": {
150
+ "video": {
151
+ "cam_high": {
152
+ "resolution": [
153
+ 640,
154
+ 480
155
+ ],
156
+ "channels": 3,
157
+ "fps": 30.0
158
+ },
159
+ "cam_left_wrist": {
160
+ "resolution": [
161
+ 640,
162
+ 480
163
+ ],
164
+ "channels": 3,
165
+ "fps": 30.0
166
+ },
167
+ "cam_right_wrist": {
168
+ "resolution": [
169
+ 640,
170
+ 480
171
+ ],
172
+ "channels": 3,
173
+ "fps": 30.0
174
+ }
175
+ },
176
+ "state": {
177
+ "right_arm": {
178
+ "absolute": true,
179
+ "rotation_type": null,
180
+ "shape": [
181
+ 6
182
+ ],
183
+ "continuous": true
184
+ },
185
+ "right_gripper": {
186
+ "absolute": true,
187
+ "rotation_type": null,
188
+ "shape": [
189
+ 1
190
+ ],
191
+ "continuous": true
192
+ }
193
+ },
194
+ "action": {
195
+ "right_arm": {
196
+ "absolute": true,
197
+ "rotation_type": null,
198
+ "shape": [
199
+ 6
200
+ ],
201
+ "continuous": true
202
+ },
203
+ "right_gripper": {
204
+ "absolute": true,
205
+ "rotation_type": null,
206
+ "shape": [
207
+ 1
208
+ ],
209
+ "continuous": true
210
+ }
211
+ }
212
+ },
213
+ "embodiment_tag": "new_embodiment"
214
+ }
215
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:278507cb3150e0c6b31d39e4c169e5a56bc69980d610698b761efa719e2c26d1
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:553d17088bb91bc355657aedae6a6837cd1946dc87cd8334830dd518e584cfea
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
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:ac86b86dfe10af973d761b55af65e944466f4c7f180c677959f81a0c3579e705
3
+ size 5713