Hugging Face's logo

float-lab
/
groot-finetune-lesandwich-combined_v2

Safetensors
gr00t_n1_5
Model card Files
xet
 Community
float-lab commited on
Commit
2458cc0
·
verified ·
1 Parent(s): 723b8f7

Upload folder using huggingface_hub

Browse files
Files changed (7) 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. 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 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 0.7119408845901489,
8
+ 1.0556365251541138,
9
+ 1.6356228590011597,
10
+ 1.6586381196975708,
11
+ 0.05677114799618721
12
+ ],
13
+ "min": [
14
+ -0.9282850027084351,
15
+ -1.8841884136199951,
16
+ -1.2719806432724,
17
+ -1.0464303493499756,
18
+ -2.122013568878174
19
+ ],
20
+ "mean": [
21
+ -0.009807169437408447,
22
+ -0.8842911124229431,
23
+ 0.9464142322540283,
24
+ 0.7901850342750549,
25
+ -1.527023196220398
26
+ ],
27
+ "std": [
28
+ 0.3438205122947693,
29
+ 0.8135552406311035,
30
+ 0.5876606106758118,
31
+ 0.4800644814968109,
32
+ 0.1549902707338333
33
+ ],
34
+ "q01": [
35
+ -0.6459636092185974,
36
+ -1.876516580581665,
37
+ -0.7579101955890655,
38
+ -0.385859699845314,
39
+ -1.778317928314209
40
+ ],
41
+ "q99": [
42
+ 0.5078716278076172,
43
+ 0.6907974624633779,
44
+ 1.6018670797348022,
45
+ 1.4234828400611874,
46
+ -1.1784463047981275
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 1.647897720336914
52
+ ],
53
+ "min": [
54
+ -0.024549685418605804
55
+ ],
56
+ "mean": [
57
+ 0.24239182472229004
58
+ ],
59
+ "std": [
60
+ 0.29583147168159485
61
+ ],
62
+ "q01": [
63
+ -0.012274842709302902
64
+ ],
65
+ "q99": [
66
+ 1.135422945022583
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 0.741093635559082,
74
+ 1.1032015085220337,
75
+ 1.6279510259628296,
76
+ 1.7691117525100708,
77
+ 0.06290856748819351
78
+ ],
79
+ "min": [
80
+ -0.9068040251731873,
81
+ -1.8903257846832275,
82
+ -1.135422945022583,
83
+ -1.1032015085220337,
84
+ -2.12508225440979
85
+ ],
86
+ "mean": [
87
+ -0.00797776598483324,
88
+ -0.9181769490242004,
89
+ 0.921265184879303,
90
+ 0.7843625545501709,
91
+ -1.5264220237731934
92
+ ],
93
+ "std": [
94
+ 0.34782350063323975,
95
+ 0.818850576877594,
96
+ 0.6179132461547852,
97
+ 0.48815080523490906,
98
+ 0.15487447381019592
99
+ ],
100
+ "q01": [
101
+ -0.6398261785507202,
102
+ -1.8857227563858032,
103
+ -0.6831256890296936,
104
+ -0.4102866220474242,
105
+ -1.779852271080017
106
+ ],
107
+ "q99": [
108
+ 0.5289536762237546,
109
+ 0.7794525623321533,
110
+ 1.6034014225006104,
111
+ 1.428085956573486,
112
+ -1.1845837640762342
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 1.799798846244812
118
+ ],
119
+ "min": [
120
+ -0.026084041222929955
121
+ ],
122
+ "mean": [
123
+ 0.23868758976459503
124
+ ],
125
+ "std": [
126
+ 0.3146224617958069
127
+ ],
128
+ "q01": [
129
+ -0.010740487836301327
130
+ ],
131
+ "q99": [
132
+ 1.2290186882019043
133
+ ]
134
+ }
135
+ }
136
+ },
137
+ "modalities": {
138
+ "video": {
139
+ "gripper_cam": {
140
+ "resolution": [
141
+ 1920,
142
+ 1080
143
+ ],
144
+ "channels": 3,
145
+ "fps": 60.0
146
+ },
147
+ "front_cam": {
148
+ "resolution": [
149
+ 1920,
150
+ 1080
151
+ ],
152
+ "channels": 3,
153
+ "fps": 60.0
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:ab10714df1e9e2e6e81d3c53404b2f03cea6b4dd6c3865421daf08852696ad06
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:f507be26f88e32c289bc8b1634866cfb1aad01be04da5da004c736ccc95df0b2
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:c03b563588b9ad938b9a34a6f45286023302458bc603c667eae0f7cd4779986d
3
+ size 5368