Hugging Face's logo

shuohsuan
/
gr00t_1m_5m

TensorBoard
Safetensors
gr00t_n1_5
Model card Files
xet
 Metrics Community
shuohsuan commited on
Commit
78e42f5
·
verified ·
1 Parent(s): 6be5ff8

Upload folder using huggingface_hub

Browse files
Files changed (10) 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/Sep24_16-21-22_dgxh00/events.out.tfevents.1758702086.dgxh00.2509029.0 +3 -0
  7. runs/Sep24_17-21-52_dgxh00/events.out.tfevents.1758705849.dgxh00.3048086.0 +3 -0
  8. runs/Sep24_18-11-04_dgxh00/events.out.tfevents.1758708698.dgxh00.2905256.0 +3 -0
  9. trainer_state.json +0 -0
  10. 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
+ 64.14427185058594,
8
+ 45.7685661315918,
9
+ 99.05348205566406,
10
+ 98.55363464355469,
11
+ 46.617828369140625
12
+ ],
13
+ "min": [
14
+ -44.20084762573242,
15
+ -99.56822204589844,
16
+ -67.15570068359375,
17
+ 23.503414154052734,
18
+ -53.5531120300293
19
+ ],
20
+ "mean": [
21
+ 6.270799275866775,
22
+ -22.805379671567486,
23
+ 32.071194584723926,
24
+ 65.53286297812282,
25
+ -13.71655155412899
26
+ ],
27
+ "std": [
28
+ 24.32212625832143,
29
+ 46.916568901665954,
30
+ 42.518874012354964,
31
+ 12.041394574375076,
32
+ 28.85272159623856
33
+ ],
34
+ "q01": [
35
+ -13.594359979877275,
36
+ -99.49402114778857,
37
+ -12.763968253443243,
38
+ 40.42789917547624,
39
+ -33.49543381997161
40
+ ],
41
+ "q99": [
42
+ 26.785697272831438,
43
+ 22.79739580448028,
44
+ 99.05348205566406,
45
+ 84.8436078092463,
46
+ 12.61398433443025
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 99.38356018066406
52
+ ],
53
+ "min": [
54
+ 0.6164383292198181
55
+ ],
56
+ "mean": [
57
+ 37.38250655007032
58
+ ],
59
+ "std": [
60
+ 36.68423457997161
61
+ ],
62
+ "q01": [
63
+ 0.7701838792146419
64
+ ],
65
+ "q99": [
66
+ 83.79630151457953
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 64.34658813476562,
74
+ 45.11668014526367,
75
+ 100.0,
76
+ 99.26078033447266,
77
+ 47.25274658203125
78
+ ],
79
+ "min": [
80
+ -44.38920593261719,
81
+ -100.0,
82
+ -71.44160461425781,
83
+ 22.135522842407227,
84
+ -53.89499282836914
85
+ ],
86
+ "mean": [
87
+ 6.2909840187044495,
88
+ -24.236251312981253,
89
+ 30.86614037801703,
90
+ 65.51736725772105,
91
+ -13.694760547376852
92
+ ],
93
+ "std": [
94
+ 24.291945723515333,
95
+ 46.31458045094955,
96
+ 43.29127625832165,
97
+ 12.21075736834071,
98
+ 28.874962228688624
99
+ ],
100
+ "q01": [
101
+ -13.609552614138392,
102
+ -99.95022232567814,
103
+ -15.089316295504398,
104
+ 39.68460520382538,
105
+ -33.63194476379998
106
+ ],
107
+ "q99": [
108
+ 26.921463318256116,
109
+ 21.350637380669006,
110
+ 99.9558022535992,
111
+ 85.06873769234312,
112
+ 12.60871293738377
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 100.0
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 35.467177022431834
124
+ ],
125
+ "std": [
126
+ 38.2207989701443
127
+ ],
128
+ "q01": [
129
+ 0.1977453265562511
130
+ ],
131
+ "q99": [
132
+ 84.46299432307067
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:b887b9a58764b1a39f70141b66f64e570b481f3e9556bf9b18cf9344e5b2dffc
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:af7727b17abb02773a1b6951e60e7ce6e183edcd38ba32c75202f6e764a42fa9
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
runs/Sep24_16-21-22_dgxh00/events.out.tfevents.1758702086.dgxh00.2509029.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9614402683ed78bf2d4a3815370c4762b7aae438a0f05ca1a222d35ea0a329d0
3
+ size 109884
runs/Sep24_17-21-52_dgxh00/events.out.tfevents.1758705849.dgxh00.3048086.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5a730abaaf8967698c8c6cb22506ab44d1acd4dad63c3d35252b045ef1ae65a
3
+ size 97272
runs/Sep24_18-11-04_dgxh00/events.out.tfevents.1758708698.dgxh00.2905256.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51996d7103aac957cb7d820077636099e4172279661af7fafc775afb9a9f45b9
3
+ size 48674
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 sha256:48602b1689d9802cfcf4207f63ff33dce3ea8808e08444baf0b0e811a2c90511
3
+ size 5368