Hugging Face's logo

lucasngoo
/
GR00T-N1.5-humanoid-pick-place-action-horizon-100

Safetensors
gr00t_n1_5
Model card Files Community
lucasngoo commited on
Commit
c63fa1a
·
verified ·
1 Parent(s): d33cfd1

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +187 -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": 100,
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": 100,
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 +1,187 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 2.2674880027770996,
8
+ 0.45470884442329407,
9
+ 8.664087295532227,
10
+ 53.32422637939453,
11
+ 25.687057495117188
12
+ ],
13
+ "min": [
14
+ -22.165273666381836,
15
+ -51.77391052246094,
16
+ -29.91075897216797,
17
+ -0.5584670901298523,
18
+ -5.630585193634033
19
+ ],
20
+ "mean": [
21
+ -4.022843360900879,
22
+ -15.838190078735352,
23
+ -19.822237014770508,
24
+ 23.733598709106445,
25
+ 3.276348114013672
26
+ ],
27
+ "std": [
28
+ 4.782382011413574,
29
+ 11.760198593139648,
30
+ 11.556901931762695,
31
+ 18.627721786499023,
32
+ 4.814205169677734
33
+ ],
34
+ "q01": [
35
+ -16.470590591430664,
36
+ -44.187110900878906,
37
+ -29.37973403930664,
38
+ -0.12512917816638947,
39
+ -4.153502464294434
40
+ ],
41
+ "q99": [
42
+ 1.2079614305496207,
43
+ -0.13733245432376862,
44
+ 0.7904205322265625,
45
+ 53.3181266784668,
46
+ 20.962854385375977
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 100.0
52
+ ],
53
+ "min": [
54
+ 0.0
55
+ ],
56
+ "mean": [
57
+ 30.3500919342041
58
+ ],
59
+ "std": [
60
+ 37.48234558105469
61
+ ],
62
+ "q01": [
63
+ 0.0
64
+ ],
65
+ "q99": [
66
+ 100.0
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 2.2674880027770996,
74
+ 0.46692728996276855,
75
+ 8.664087295532227,
76
+ 53.336429595947266,
77
+ 25.687057495117188
78
+ ],
79
+ "min": [
80
+ -22.165273666381836,
81
+ -51.77391052246094,
82
+ -29.91075897216797,
83
+ -0.5584670901298523,
84
+ -5.630585193634033
85
+ ],
86
+ "mean": [
87
+ -4.022891044616699,
88
+ -15.825581550598145,
89
+ -19.828760147094727,
90
+ 23.717830657958984,
91
+ 3.2795426845550537
92
+ ],
93
+ "std": [
94
+ 4.781561374664307,
95
+ 11.76130199432373,
96
+ 11.552718162536621,
97
+ 18.630279541015625,
98
+ 4.816193103790283
99
+ ],
100
+ "q01": [
101
+ -16.470590591430664,
102
+ -44.167692527770996,
103
+ -29.37973403930664,
104
+ -0.12512917816638947,
105
+ -4.153502464294434
106
+ ],
107
+ "q99": [
108
+ 1.211570382118225,
109
+ -0.13733245432376862,
110
+ 0.7904205322265625,
111
+ 53.3181266784668,
112
+ 20.962854385375977
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 100.0
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 30.33521842956543
124
+ ],
125
+ "std": [
126
+ 37.47881317138672
127
+ ],
128
+ "q01": [
129
+ 0.0
130
+ ],
131
+ "q99": [
132
+ 100.0
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
+ },
148
+ "state": {
149
+ "single_arm": {
150
+ "absolute": true,
151
+ "rotation_type": null,
152
+ "shape": [
153
+ 5
154
+ ],
155
+ "continuous": true
156
+ },
157
+ "gripper": {
158
+ "absolute": true,
159
+ "rotation_type": null,
160
+ "shape": [
161
+ 1
162
+ ],
163
+ "continuous": true
164
+ }
165
+ },
166
+ "action": {
167
+ "single_arm": {
168
+ "absolute": true,
169
+ "rotation_type": null,
170
+ "shape": [
171
+ 5
172
+ ],
173
+ "continuous": true
174
+ },
175
+ "gripper": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 1
180
+ ],
181
+ "continuous": true
182
+ }
183
+ }
184
+ },
185
+ "embodiment_tag": "new_embodiment"
186
+ }
187
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:535e3ee2cabf65b32541dc445d3652703b586bfa3ebb81a06fa757d33168f9a6
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36c6311003a0adaf0d25dceefc1121b812c6f88c87932516656924d9a2a075d5
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:a2f03cb555b119a039b71f4bd58704e9747f10ab5985f65ce347d6d42a139461
3
+ size 8550720062
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24aa855818aaa4be28c45d149716fa4746afa632987fdf97f293aa742b3a393f
3
+ size 14244
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96dedd731c018af5121e5a6273b315a5ea44ce825ed023675cb4a9716b4911be
3
+ size 1064
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff