Hugging Face's logo

shaunkirby
/
multi-task

Safetensors
gr00t_n1_5
Model card Files
xet
Community
shaunkirby commited on
Commit
5143014
·
verified ·
1 Parent(s): d00c7f2

Upload folder using huggingface_hub

Browse files
Files changed (8) 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. scheduler.pt +3 -0
  8. 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
+ 13.750532150268555,
8
+ 65.29716491699219,
9
+ 99.72948455810547,
10
+ 98.54327392578125,
11
+ 56.414947509765625
12
+ ],
13
+ "min": [
14
+ -72.4137954711914,
15
+ -98.32469177246094,
16
+ -95.76194763183594,
17
+ -36.675235748291016,
18
+ -46.90357971191406
19
+ ],
20
+ "mean": [
21
+ -19.597606658935547,
22
+ -39.21821594238281,
23
+ 46.42843246459961,
24
+ 67.89632415771484,
25
+ 9.61771297454834
26
+ ],
27
+ "std": [
28
+ 13.64588451385498,
29
+ 53.19573974609375,
30
+ 54.47523880004883,
31
+ 13.680619239807129,
32
+ 13.932531356811523
33
+ ],
34
+ "q01": [
35
+ -60.7492561340332,
36
+ -98.08535766601562,
37
+ -78.43552513122559,
38
+ 45.92973327636719,
39
+ -35.928924560546875
40
+ ],
41
+ "q99": [
42
+ 5.576841354370117,
43
+ 53.011566162109375,
44
+ 99.27863311767578,
45
+ 98.45758056640625,
46
+ 47.217140197753906
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 38.30351257324219
52
+ ],
53
+ "min": [
54
+ 1.789264440536499
55
+ ],
56
+ "mean": [
57
+ 8.25131893157959
58
+ ],
59
+ "std": [
60
+ 8.493812561035156
61
+ ],
62
+ "q01": [
63
+ 1.789264440536499
64
+ ],
65
+ "q99": [
66
+ 33.59840774536133
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 13.954358100891113,
74
+ 63.529903411865234,
75
+ 100.0,
76
+ 100.0,
77
+ 56.702659606933594
78
+ ],
79
+ "min": [
80
+ -72.68986511230469,
81
+ -100.0,
82
+ -99.63652801513672,
83
+ -37.81584548950195,
84
+ -47.27416229248047
85
+ ],
86
+ "mean": [
87
+ -19.6429386138916,
88
+ -41.03962707519531,
89
+ 44.71687698364258,
90
+ 67.806640625,
91
+ 9.598431587219238
92
+ ],
93
+ "std": [
94
+ 13.604964256286621,
95
+ 52.89664077758789,
96
+ 55.28606414794922,
97
+ 14.091605186462402,
98
+ 13.937583923339844
99
+ ],
100
+ "q01": [
101
+ -60.84549198150634,
102
+ -100.0,
103
+ -81.7355728149414,
104
+ 45.09635925292969,
105
+ -36.04687805175781
106
+ ],
107
+ "q99": [
108
+ 5.8735504150390625,
109
+ 50.64826583862305,
110
+ 100.0,
111
+ 99.65738677978516,
112
+ 47.010799407958984
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 47.049312591552734
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 6.361488342285156
124
+ ],
125
+ "std": [
126
+ 9.334465980529785
127
+ ],
128
+ "q01": [
129
+ 0.5658851861953735
130
+ ],
131
+ "q99": [
132
+ 34.27647399902344
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "side": {
140
+ "resolution": [
141
+ 1280,
142
+ 720
143
+ ],
144
+ "channels": 3,
145
+ "fps": 30.0
146
+ },
147
+ "top": {
148
+ "resolution": [
149
+ 1280,
150
+ 720
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:b456c7555f0755a3ec82014aa09b5f42f7e30568336d2fb9a085dbeabdab1afa
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:cac92336b1cdc6f8aca691350ba49c31fed1d7bfeb37d2f90d17f595fdf1df63
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 sha256:7f112dfe4efb42442927ad4a966795896fc75ccfcc9510f4cc47943e453c8b85
3
+ size 8550720062
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