Hugging Face's logo

siyulw2025
/
gr00t_orange_cache

Safetensors
Model card Files
xet
Community
siyulw2025 commited on
Commit
c757934
·
verified ·
1 Parent(s): 32de4d4

Upload Groot model checkpoint

Browse files
Files changed (10) hide show
  1. checkpoint-10000/config.json +64 -0
  2. checkpoint-10000/experiment_cfg/metadata.json +195 -0
  3. checkpoint-10000/model-00001-of-00002.safetensors +3 -0
  4. checkpoint-10000/model-00002-of-00002.safetensors +3 -0
  5. checkpoint-10000/model.safetensors.index.json +0 -0
  6. checkpoint-10000/optimizer.pt +3 -0
  7. checkpoint-10000/rng_state.pth +3 -0
  8. checkpoint-10000/scheduler.pt +3 -0
  9. checkpoint-10000/trainer_state.json +0 -0
  10. experiment_cfg/metadata.json +195 -0
checkpoint-10000/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
+ }
checkpoint-10000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 70.94871520996094,
8
+ 77.22320556640625,
9
+ 100.01800537109375,
10
+ 100.16673278808594,
11
+ 97.60052490234375
12
+ ],
13
+ "min": [
14
+ -69.23458862304688,
15
+ -100.0243911743164,
16
+ -73.13743591308594,
17
+ -30.78144073486328,
18
+ -98.89419555664062
19
+ ],
20
+ "mean": [
21
+ 9.360292434692383,
22
+ -16.005531311035156,
23
+ 34.987754821777344,
24
+ 37.55447769165039,
25
+ -72.28470611572266
26
+ ],
27
+ "std": [
28
+ 20.799467086791992,
29
+ 41.01473617553711,
30
+ 36.12113571166992,
31
+ 23.66265106201172,
32
+ 12.978203773498535
33
+ ],
34
+ "q01": [
35
+ -35.997765617370604,
36
+ -100.0104158782959,
37
+ -46.288869018554685,
38
+ -13.32153076171875,
39
+ -94.14221015930175
40
+ ],
41
+ "q99": [
42
+ 46.8429727172852,
43
+ 52.10946578979511,
44
+ 100.00050704956055,
45
+ 87.94155975341809,
46
+ -29.841210021972657
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 95.22621154785156
52
+ ],
53
+ "min": [
54
+ -0.2612617015838623
55
+ ],
56
+ "mean": [
57
+ 35.83467483520508
58
+ ],
59
+ "std": [
60
+ 18.614166259765625
61
+ ],
62
+ "q01": [
63
+ 0.0655644954741001
64
+ ],
65
+ "q99": [
66
+ 73.44229194641116
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 75.80465698242188,
74
+ 74.10256958007812,
75
+ 100.0,
76
+ 100.0,
77
+ 100.0
78
+ ],
79
+ "min": [
80
+ -70.10728454589844,
81
+ -99.99999237060547,
82
+ -71.726318359375,
83
+ -31.92371368408203,
84
+ -99.95116424560547
85
+ ],
86
+ "mean": [
87
+ 9.064373016357422,
88
+ -17.50950813293457,
89
+ 34.298583984375,
90
+ 35.80142593383789,
91
+ -72.71471405029297
92
+ ],
93
+ "std": [
94
+ 20.691120147705078,
95
+ 41.16417694091797,
96
+ 36.993385314941406,
97
+ 25.210084915161133,
98
+ 17.97390365600586
99
+ ],
100
+ "q01": [
101
+ -36.475023727416996,
102
+ -99.82906341552734,
103
+ -47.98368835449219,
104
+ -16.169151306152344,
105
+ -95.31135559082031
106
+ ],
107
+ "q99": [
108
+ 46.35589599609375,
109
+ 49.23077392578125,
110
+ 100.0,
111
+ 88.41044342041042,
112
+ -43.83394241333008
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 95.49867248535156
118
+ ],
119
+ "min": [
120
+ 8.669766202729079e-07
121
+ ],
122
+ "mean": [
123
+ 25.232954025268555
124
+ ],
125
+ "std": [
126
+ 25.212261199951172
127
+ ],
128
+ "q01": [
129
+ 8.669766202729079e-07
130
+ ],
131
+ "q99": [
132
+ 74.31597137451172
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
+ "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
+ }
checkpoint-10000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19ae89b3f6440c0c0f0e78b9858436c8a5fad59e95a6c637e0df2ab7a496b340
3
+ size 4999367032
checkpoint-10000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75a4f075db9434862bbe8b0022ece444a1ad0370a2ce7a3ac74298216667e7b3
3
+ size 2586705312
checkpoint-10000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4df7995908b105ced17e14e5a83f499a22a3e9b97e179bdb0bf348898f6a986
3
+ size 8550720062
checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8985307d5f2396161858ee2ff90079044e62e63df5af802a7bd5d83237b7ddc4
3
+ size 14244
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6720e2e5ea7a31d46acf7bf43755c1d4488376f4ec21c9257477d1d473150671
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 70.94871520996094,
8
+ 77.22320556640625,
9
+ 100.01800537109375,
10
+ 100.16673278808594,
11
+ 97.60052490234375
12
+ ],
13
+ "min": [
14
+ -69.23458862304688,
15
+ -100.0243911743164,
16
+ -73.13743591308594,
17
+ -30.78144073486328,
18
+ -98.89419555664062
19
+ ],
20
+ "mean": [
21
+ 9.360292434692383,
22
+ -16.005531311035156,
23
+ 34.987754821777344,
24
+ 37.55447769165039,
25
+ -72.28470611572266
26
+ ],
27
+ "std": [
28
+ 20.799467086791992,
29
+ 41.01473617553711,
30
+ 36.12113571166992,
31
+ 23.66265106201172,
32
+ 12.978203773498535
33
+ ],
34
+ "q01": [
35
+ -35.997765617370604,
36
+ -100.0104158782959,
37
+ -46.288869018554685,
38
+ -13.32153076171875,
39
+ -94.14221015930175
40
+ ],
41
+ "q99": [
42
+ 46.8429727172852,
43
+ 52.10946578979511,
44
+ 100.00050704956055,
45
+ 87.94155975341809,
46
+ -29.841210021972657
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 95.22621154785156
52
+ ],
53
+ "min": [
54
+ -0.2612617015838623
55
+ ],
56
+ "mean": [
57
+ 35.83467483520508
58
+ ],
59
+ "std": [
60
+ 18.614166259765625
61
+ ],
62
+ "q01": [
63
+ 0.0655644954741001
64
+ ],
65
+ "q99": [
66
+ 73.44229194641116
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 75.80465698242188,
74
+ 74.10256958007812,
75
+ 100.0,
76
+ 100.0,
77
+ 100.0
78
+ ],
79
+ "min": [
80
+ -70.10728454589844,
81
+ -99.99999237060547,
82
+ -71.726318359375,
83
+ -31.92371368408203,
84
+ -99.95116424560547
85
+ ],
86
+ "mean": [
87
+ 9.064373016357422,
88
+ -17.50950813293457,
89
+ 34.298583984375,
90
+ 35.80142593383789,
91
+ -72.71471405029297
92
+ ],
93
+ "std": [
94
+ 20.691120147705078,
95
+ 41.16417694091797,
96
+ 36.993385314941406,
97
+ 25.210084915161133,
98
+ 17.97390365600586
99
+ ],
100
+ "q01": [
101
+ -36.475023727416996,
102
+ -99.82906341552734,
103
+ -47.98368835449219,
104
+ -16.169151306152344,
105
+ -95.31135559082031
106
+ ],
107
+ "q99": [
108
+ 46.35589599609375,
109
+ 49.23077392578125,
110
+ 100.0,
111
+ 88.41044342041042,
112
+ -43.83394241333008
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 95.49867248535156
118
+ ],
119
+ "min": [
120
+ 8.669766202729079e-07
121
+ ],
122
+ "mean": [
123
+ 25.232954025268555
124
+ ],
125
+ "std": [
126
+ 25.212261199951172
127
+ ],
128
+ "q01": [
129
+ 8.669766202729079e-07
130
+ ],
131
+ "q99": [
132
+ 74.31597137451172
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
+ "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
+ }