Hugging Face's logo

shuohsuan
/
gr00t_135a_24m

TensorBoard
Safetensors
gr00t_n1_5
Model card Files
xet
Metrics Community
shuohsuan commited on
Commit
444395c
·
verified ·
1 Parent(s): 2fff784

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. runs/Sep25_10-48-26_dgxh00/events.out.tfevents.1758768512.dgxh00.951639.0 +3 -0
  7. runs/Sep25_15-50-18_dgxh00/events.out.tfevents.1758786754.dgxh00.3367626.0 +3 -0
  8. trainer_state.json +0 -0
  9. 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 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 64.21499633789062,
8
+ 36.096717834472656,
9
+ 99.14813232421875,
10
+ 94.37525177001953,
11
+ 45.64102554321289
12
+ ],
13
+ "min": [
14
+ -43.069305419921875,
15
+ -99.56822204589844,
16
+ -26.265972137451172,
17
+ -16.99477767944336,
18
+ -52.380950927734375
19
+ ],
20
+ "mean": [
21
+ 5.509986158382245,
22
+ -14.35335840673113,
23
+ 26.20888646459224,
24
+ 64.48625555692769,
25
+ -14.555836104261807
26
+ ],
27
+ "std": [
28
+ 24.187165371505877,
29
+ 43.93993290498261,
30
+ 38.391289630336004,
31
+ 12.210329297075871,
32
+ 28.946839681941185
33
+ ],
34
+ "q01": [
35
+ -13.477248074552511,
36
+ -99.42778812114173,
37
+ -11.921842661227698,
38
+ 33.67408155724538,
39
+ -34.081507537501416
40
+ ],
41
+ "q99": [
42
+ 26.429990228537005,
43
+ 23.947844406312953,
44
+ 99.05348205566403,
45
+ 86.4698453432226,
46
+ 13.711661886088773
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 99.38356018066406
52
+ ],
53
+ "min": [
54
+ 0.6849315166473389
55
+ ],
56
+ "mean": [
57
+ 36.31173445260529
58
+ ],
59
+ "std": [
60
+ 33.42170779284523
61
+ ],
62
+ "q01": [
63
+ 0.7492336264384035
64
+ ],
65
+ "q99": [
66
+ 87.56269824034457
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 64.34658813476562,
74
+ 35.263614654541016,
75
+ 100.0,
76
+ 94.8254623413086,
77
+ 46.227108001708984
78
+ ],
79
+ "min": [
80
+ -43.181819915771484,
81
+ -100.0,
82
+ -29.105838775634766,
83
+ -18.767967224121094,
84
+ -52.72283172607422
85
+ ],
86
+ "mean": [
87
+ 5.475849194900448,
88
+ -15.916569921625413,
89
+ 24.91718032942351,
90
+ 64.43485916360095,
91
+ -14.53409389601117
92
+ ],
93
+ "std": [
94
+ 24.157554977049045,
95
+ 43.455208385946506,
96
+ 39.16846289161519,
97
+ 12.375006245412076,
98
+ 28.9809541700346
99
+ ],
100
+ "q01": [
101
+ -13.41717684960592,
102
+ -99.94738485191877,
103
+ -14.348397427466344,
104
+ 32.821429070434746,
105
+ -34.151134127281416
106
+ ],
107
+ "q99": [
108
+ 26.408953742903826,
109
+ 22.903307667960902,
110
+ 99.95137978888387,
111
+ 86.63989080352458,
112
+ 13.933328256339754
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 100.0
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 34.37169643936741
124
+ ],
125
+ "std": [
126
+ 35.016087209431035
127
+ ],
128
+ "q01": [
129
+ 0.1746102333150955
130
+ ],
131
+ "q99": [
132
+ 88.09365581932163
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "side": {
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 sha256:1426e4a403e4ce91ef398cf4a797f407808fc53323c59f23a46f0d30431e5697
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ec91df2a1d37704bf0889322c95b60182e6f37ff475b4ddf11a9a6820b981d6
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
runs/Sep25_10-48-26_dgxh00/events.out.tfevents.1758768512.dgxh00.951639.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1227eea7f07749eea2f895d33737beb360689c17248bb93d74b9f7ab83ea0540
3
+ size 81408
runs/Sep25_15-50-18_dgxh00/events.out.tfevents.1758786754.dgxh00.3367626.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2df937c8b68f2f3c32b32dcc30c5b481b198ad05c21fe516b3ad0c0fcab9c88
3
+ size 154183
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:ca55870b183cdf81ed2fd2c30355ed81fb7a53413a51eb47291ceb67894fa4b3
3
+ size 5368