Hugging Face's logo

binhng
/
gr15_open_pot_BASE

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
binhng commited on
Commit
0e8063c
·
verified ·
1 Parent(s): 67f4a21

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. config.json +64 -0
  2. experiment_cfg/metadata.json +207 -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": "4.51.3"
64
+ }
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "joint_state": {
6
+ "max": [
7
+ 3.35524320602417,
8
+ -1.0901570320129395,
9
+ 1.921777606010437,
10
+ -1.7477141618728638,
11
+ -1.486493468284607,
12
+ 0.28720736503601074
13
+ ],
14
+ "min": [
15
+ 2.8941030502319336,
16
+ -1.5188004970550537,
17
+ 1.5325502157211304,
18
+ -2.3195159435272217,
19
+ -1.8238309621810913,
20
+ -0.3950203061103821
21
+ ],
22
+ "mean": [
23
+ 3.117759943008423,
24
+ -1.2923474311828613,
25
+ 1.7299931049346924,
26
+ -1.9779006242752075,
27
+ -1.6687124967575073,
28
+ -0.008849170990288258
29
+ ],
30
+ "std": [
31
+ 0.11333384364843369,
32
+ 0.07360938936471939,
33
+ 0.06820472329854965,
34
+ 0.0754924789071083,
35
+ 0.05238862335681915,
36
+ 0.11165500432252884
37
+ ],
38
+ "q01": [
39
+ 2.919258089065552,
40
+ -1.4639486396312713,
41
+ 1.5665747356414794,
42
+ -2.1587486171722414,
43
+ -1.786708240509033,
44
+ -0.27216118574142456
45
+ ],
46
+ "q99": [
47
+ 3.3308198475837707,
48
+ -1.134760583639145,
49
+ 1.8730972576141356,
50
+ -1.826737403869629,
51
+ -1.5515401363372803,
52
+ 0.22390946477651594
53
+ ]
54
+ },
55
+ "gripper": {
56
+ "max": [
57
+ 228.0
58
+ ],
59
+ "min": [
60
+ 3.0
61
+ ],
62
+ "mean": [
63
+ 171.0463104248047
64
+ ],
65
+ "std": [
66
+ 93.9524154663086
67
+ ],
68
+ "q01": [
69
+ 3.0
70
+ ],
71
+ "q99": [
72
+ 228.0
73
+ ]
74
+ }
75
+ },
76
+ "action": {
77
+ "joint_state": {
78
+ "max": [
79
+ 3.35524320602417,
80
+ -1.0901570320129395,
81
+ 1.921777606010437,
82
+ -1.7477141618728638,
83
+ -1.486493468284607,
84
+ 0.28720736503601074
85
+ ],
86
+ "min": [
87
+ 2.8941030502319336,
88
+ -1.5188004970550537,
89
+ 1.5325502157211304,
90
+ -2.3195159435272217,
91
+ -1.8238309621810913,
92
+ -0.3950203061103821
93
+ ],
94
+ "mean": [
95
+ 3.117722749710083,
96
+ -1.29246985912323,
97
+ 1.7300474643707275,
98
+ -1.9778891801834106,
99
+ -1.6687861680984497,
100
+ -0.008841215632855892
101
+ ],
102
+ "std": [
103
+ 0.11333613097667694,
104
+ 0.07367856800556183,
105
+ 0.06822366267442703,
106
+ 0.07544912397861481,
107
+ 0.05228527635335922,
108
+ 0.11165928095579147
109
+ ],
110
+ "q01": [
111
+ 2.919258089065552,
112
+ -1.464362633228302,
113
+ 1.5665747356414794,
114
+ -2.1587486171722414,
115
+ -1.7864461660385131,
116
+ -0.27216118574142456
117
+ ],
118
+ "q99": [
119
+ 3.3308198475837707,
120
+ -1.134760583639145,
121
+ 1.873173861503601,
122
+ -1.8267549419403075,
123
+ -1.5515621089935303,
124
+ 0.22390946477651594
125
+ ]
126
+ },
127
+ "gripper": {
128
+ "max": [
129
+ 1.0
130
+ ],
131
+ "min": [
132
+ -1.0
133
+ ],
134
+ "mean": [
135
+ -0.5914595127105713
136
+ ],
137
+ "std": [
138
+ 0.8063073754310608
139
+ ],
140
+ "q01": [
141
+ -1.0
142
+ ],
143
+ "q99": [
144
+ 1.0
145
+ ]
146
+ }
147
+ }
148
+ },
149
+ "modalities": {
150
+ "video": {
151
+ "image": {
152
+ "resolution": [
153
+ 720,
154
+ 540
155
+ ],
156
+ "channels": 3,
157
+ "fps": 20.0
158
+ },
159
+ "wrist_image": {
160
+ "resolution": [
161
+ 720,
162
+ 540
163
+ ],
164
+ "channels": 3,
165
+ "fps": 20.0
166
+ }
167
+ },
168
+ "state": {
169
+ "joint_state": {
170
+ "absolute": true,
171
+ "rotation_type": null,
172
+ "shape": [
173
+ 6
174
+ ],
175
+ "continuous": true
176
+ },
177
+ "gripper": {
178
+ "absolute": true,
179
+ "rotation_type": null,
180
+ "shape": [
181
+ 1
182
+ ],
183
+ "continuous": true
184
+ }
185
+ },
186
+ "action": {
187
+ "joint_state": {
188
+ "absolute": true,
189
+ "rotation_type": null,
190
+ "shape": [
191
+ 6
192
+ ],
193
+ "continuous": true
194
+ },
195
+ "gripper": {
196
+ "absolute": true,
197
+ "rotation_type": null,
198
+ "shape": [
199
+ 1
200
+ ],
201
+ "continuous": true
202
+ }
203
+ }
204
+ },
205
+ "embodiment_tag": "new_embodiment"
206
+ }
207
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce235eba7f53826d208b78d6de33f103c153ccf37f3a9cee8dc6d40f54b7ebaf
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c1dd100caa5fa079bb4989c83f620c44c72a57a311494545c12a819d71ee0c5
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 sha256:5f3446cffed9c41bf1985be3d347f676182af720a1fb60e24af6d5159f151644
3
+ size 5560