Hugging Face's logo

nimitt0
/
pick_objects_multiscene30k

Safetensors
gr00t_n1_5
Model card Files
xet
Community
nimitt0 commited on
Commit
40990a7
·
verified ·
1 Parent(s): 30dad99

Add files using upload-large-folder tool

Browse files
Files changed (34) 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. checkpoint-20000/config.json +64 -0
  11. checkpoint-20000/experiment_cfg/metadata.json +195 -0
  12. checkpoint-20000/model-00001-of-00002.safetensors +3 -0
  13. checkpoint-20000/model-00002-of-00002.safetensors +3 -0
  14. checkpoint-20000/model.safetensors.index.json +0 -0
  15. checkpoint-20000/optimizer.pt +3 -0
  16. checkpoint-20000/rng_state.pth +3 -0
  17. checkpoint-20000/scheduler.pt +3 -0
  18. checkpoint-20000/trainer_state.json +0 -0
  19. checkpoint-30000/config.json +64 -0
  20. checkpoint-30000/experiment_cfg/metadata.json +195 -0
  21. checkpoint-30000/model-00001-of-00002.safetensors +3 -0
  22. checkpoint-30000/model-00002-of-00002.safetensors +3 -0
  23. checkpoint-30000/model.safetensors.index.json +0 -0
  24. checkpoint-30000/optimizer.pt +3 -0
  25. checkpoint-30000/rng_state.pth +3 -0
  26. checkpoint-30000/scheduler.pt +3 -0
  27. checkpoint-30000/trainer_state.json +0 -0
  28. config.json +64 -0
  29. experiment_cfg/metadata.json +195 -0
  30. model-00001-of-00002.safetensors +3 -0
  31. model-00002-of-00002.safetensors +3 -0
  32. model.safetensors.index.json +0 -0
  33. trainer_state.json +0 -0
  34. training_args.bin +3 -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
+ 1.6115039587020874,
8
+ 3.1076648235321045,
9
+ 0.2685627043247223,
10
+ 1.6672509908676147,
11
+ 1.969820261001587
12
+ ],
13
+ "min": [
14
+ -1.908600926399231,
15
+ -0.17690595984458923,
16
+ -2.865432024002075,
17
+ -2.654787063598633,
18
+ -1.9672902822494507
19
+ ],
20
+ "mean": [
21
+ -0.00959783991910409,
22
+ 1.499808118712209,
23
+ -1.2723766745625318,
24
+ 0.6897374995980092,
25
+ 0.012068234835772905
26
+ ],
27
+ "std": [
28
+ 0.8033180848336506,
29
+ 0.6881857246367352,
30
+ 0.5993451999991707,
31
+ 0.47767480027308595,
32
+ 0.8412626329126346
33
+ ],
34
+ "q01": [
35
+ -1.4122188767992025,
36
+ 0.04034731964515659,
37
+ -2.793325455606157,
38
+ -0.9005685478050315,
39
+ -1.4498776078523188
40
+ ],
41
+ "q99": [
42
+ 1.4172475041389343,
43
+ 2.8561244349134496,
44
+ -0.21244733821766565,
45
+ 1.2000015139915712,
46
+ 1.4504311649198878
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 1.9673068523406982
52
+ ],
53
+ "min": [
54
+ -0.03646281734108925
55
+ ],
56
+ "mean": [
57
+ 0.4539852668266194
58
+ ],
59
+ "std": [
60
+ 0.31508159664095936
61
+ ],
62
+ "q01": [
63
+ 0.05382254484269338
64
+ ],
65
+ "q99": [
66
+ 1.4516974584290336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 1.6115039587020874,
74
+ 3.1076648235321045,
75
+ 0.24772115051746368,
76
+ 1.6672509908676147,
77
+ 1.9685454368591309
78
+ ],
79
+ "min": [
80
+ -1.908600926399231,
81
+ -0.11258061230182648,
82
+ -2.865432024002075,
83
+ -2.65680193901062,
84
+ -1.9372831583023071
85
+ ],
86
+ "mean": [
87
+ -0.00968967524832796,
88
+ 1.506859118246074,
89
+ -1.2779861339022518,
90
+ 0.6942818922168711,
91
+ 0.012125308062864543
92
+ ],
93
+ "std": [
94
+ 0.8053480669448868,
95
+ 0.6855857436916613,
96
+ 0.5989715373901173,
97
+ 0.47624700980912044,
98
+ 0.8405142018342715
99
+ ],
100
+ "q01": [
101
+ -1.4130165935286558,
102
+ 0.044443692726112066,
103
+ -2.7953697018785206,
104
+ -0.8843341536846144,
105
+ -1.4473891171261197
106
+ ],
107
+ "q99": [
108
+ 1.417625653380196,
109
+ 2.8571220874299215,
110
+ -0.2219349648594896,
111
+ 1.2000015139915712,
112
+ 1.4479686097661741
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 1.937300443649292
118
+ ],
119
+ "min": [
120
+ -0.03744976595044136
121
+ ],
122
+ "mean": [
123
+ 0.4494256477809244
124
+ ],
125
+ "std": [
126
+ 0.3115188437483966
127
+ ],
128
+ "q01": [
129
+ 0.05372234335522662
130
+ ],
131
+ "q99": [
132
+ 1.42285219032216
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 sha256:8fe5ae06918cadd78b2d233df9060d65e0f20b3fe59906a776aa36a7cfead4f3
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 sha256:9524ba30093e2621251c579b2da4b084bba5403d3534afbbbcd1144e33911742
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:a7380bd6614965115023f031c4a562497b36cce44eaf959a3dd45cdda738dede
3
+ size 10210298785
checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c58da6049dabbab024b5bf3f0963d97425721855f521b80b177eaf1c40192c1
3
+ size 14244
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69cc8a263533866a30bbd70c695b627ba4dea5733aa097b95dcfa6e7f21407ca
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/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-20000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 1.6115039587020874,
8
+ 3.1076648235321045,
9
+ 0.2685627043247223,
10
+ 1.6672509908676147,
11
+ 1.969820261001587
12
+ ],
13
+ "min": [
14
+ -1.908600926399231,
15
+ -0.17690595984458923,
16
+ -2.865432024002075,
17
+ -2.654787063598633,
18
+ -1.9672902822494507
19
+ ],
20
+ "mean": [
21
+ -0.00959783991910409,
22
+ 1.499808118712209,
23
+ -1.2723766745625318,
24
+ 0.6897374995980092,
25
+ 0.012068234835772905
26
+ ],
27
+ "std": [
28
+ 0.8033180848336506,
29
+ 0.6881857246367352,
30
+ 0.5993451999991707,
31
+ 0.47767480027308595,
32
+ 0.8412626329126346
33
+ ],
34
+ "q01": [
35
+ -1.4122188767992025,
36
+ 0.04034731964515659,
37
+ -2.793325455606157,
38
+ -0.9005685478050315,
39
+ -1.4498776078523188
40
+ ],
41
+ "q99": [
42
+ 1.4172475041389343,
43
+ 2.8561244349134496,
44
+ -0.21244733821766565,
45
+ 1.2000015139915712,
46
+ 1.4504311649198878
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 1.9673068523406982
52
+ ],
53
+ "min": [
54
+ -0.03646281734108925
55
+ ],
56
+ "mean": [
57
+ 0.4539852668266194
58
+ ],
59
+ "std": [
60
+ 0.31508159664095936
61
+ ],
62
+ "q01": [
63
+ 0.05382254484269338
64
+ ],
65
+ "q99": [
66
+ 1.4516974584290336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 1.6115039587020874,
74
+ 3.1076648235321045,
75
+ 0.24772115051746368,
76
+ 1.6672509908676147,
77
+ 1.9685454368591309
78
+ ],
79
+ "min": [
80
+ -1.908600926399231,
81
+ -0.11258061230182648,
82
+ -2.865432024002075,
83
+ -2.65680193901062,
84
+ -1.9372831583023071
85
+ ],
86
+ "mean": [
87
+ -0.00968967524832796,
88
+ 1.506859118246074,
89
+ -1.2779861339022518,
90
+ 0.6942818922168711,
91
+ 0.012125308062864543
92
+ ],
93
+ "std": [
94
+ 0.8053480669448868,
95
+ 0.6855857436916613,
96
+ 0.5989715373901173,
97
+ 0.47624700980912044,
98
+ 0.8405142018342715
99
+ ],
100
+ "q01": [
101
+ -1.4130165935286558,
102
+ 0.044443692726112066,
103
+ -2.7953697018785206,
104
+ -0.8843341536846144,
105
+ -1.4473891171261197
106
+ ],
107
+ "q99": [
108
+ 1.417625653380196,
109
+ 2.8571220874299215,
110
+ -0.2219349648594896,
111
+ 1.2000015139915712,
112
+ 1.4479686097661741
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 1.937300443649292
118
+ ],
119
+ "min": [
120
+ -0.03744976595044136
121
+ ],
122
+ "mean": [
123
+ 0.4494256477809244
124
+ ],
125
+ "std": [
126
+ 0.3115188437483966
127
+ ],
128
+ "q01": [
129
+ 0.05372234335522662
130
+ ],
131
+ "q99": [
132
+ 1.42285219032216
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-20000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dab843e59f73a1619fa7a063efe283472efd61130895aa9591f513df4749c479
3
+ size 4999367032
checkpoint-20000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b593f1ff8dc8be892ff818b5ae71f6e3b651a241b7bd2e2e298bac67d9ea39c
3
+ size 2586705312
checkpoint-20000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec03ea9c1d5ce6fa5fc7a2df7f150bf1695086f5051473a49e6689f8b87ed6e1
3
+ size 10210298785
checkpoint-20000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1e423e01daba057dc48f9ccf1a4c9e5ea7f2124241d644fb58e681c20997e0d
3
+ size 14244
checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:514333fcc99a00a5ac2f0aadb8529bebf896b0c8433955834346694f8797b343
3
+ size 1064
checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/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-30000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 1.6115039587020874,
8
+ 3.1076648235321045,
9
+ 0.2685627043247223,
10
+ 1.6672509908676147,
11
+ 1.969820261001587
12
+ ],
13
+ "min": [
14
+ -1.908600926399231,
15
+ -0.17690595984458923,
16
+ -2.865432024002075,
17
+ -2.654787063598633,
18
+ -1.9672902822494507
19
+ ],
20
+ "mean": [
21
+ -0.00959783991910409,
22
+ 1.499808118712209,
23
+ -1.2723766745625318,
24
+ 0.6897374995980092,
25
+ 0.012068234835772905
26
+ ],
27
+ "std": [
28
+ 0.8033180848336506,
29
+ 0.6881857246367352,
30
+ 0.5993451999991707,
31
+ 0.47767480027308595,
32
+ 0.8412626329126346
33
+ ],
34
+ "q01": [
35
+ -1.4122188767992025,
36
+ 0.04034731964515659,
37
+ -2.793325455606157,
38
+ -0.9005685478050315,
39
+ -1.4498776078523188
40
+ ],
41
+ "q99": [
42
+ 1.4172475041389343,
43
+ 2.8561244349134496,
44
+ -0.21244733821766565,
45
+ 1.2000015139915712,
46
+ 1.4504311649198878
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 1.9673068523406982
52
+ ],
53
+ "min": [
54
+ -0.03646281734108925
55
+ ],
56
+ "mean": [
57
+ 0.4539852668266194
58
+ ],
59
+ "std": [
60
+ 0.31508159664095936
61
+ ],
62
+ "q01": [
63
+ 0.05382254484269338
64
+ ],
65
+ "q99": [
66
+ 1.4516974584290336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 1.6115039587020874,
74
+ 3.1076648235321045,
75
+ 0.24772115051746368,
76
+ 1.6672509908676147,
77
+ 1.9685454368591309
78
+ ],
79
+ "min": [
80
+ -1.908600926399231,
81
+ -0.11258061230182648,
82
+ -2.865432024002075,
83
+ -2.65680193901062,
84
+ -1.9372831583023071
85
+ ],
86
+ "mean": [
87
+ -0.00968967524832796,
88
+ 1.506859118246074,
89
+ -1.2779861339022518,
90
+ 0.6942818922168711,
91
+ 0.012125308062864543
92
+ ],
93
+ "std": [
94
+ 0.8053480669448868,
95
+ 0.6855857436916613,
96
+ 0.5989715373901173,
97
+ 0.47624700980912044,
98
+ 0.8405142018342715
99
+ ],
100
+ "q01": [
101
+ -1.4130165935286558,
102
+ 0.044443692726112066,
103
+ -2.7953697018785206,
104
+ -0.8843341536846144,
105
+ -1.4473891171261197
106
+ ],
107
+ "q99": [
108
+ 1.417625653380196,
109
+ 2.8571220874299215,
110
+ -0.2219349648594896,
111
+ 1.2000015139915712,
112
+ 1.4479686097661741
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 1.937300443649292
118
+ ],
119
+ "min": [
120
+ -0.03744976595044136
121
+ ],
122
+ "mean": [
123
+ 0.4494256477809244
124
+ ],
125
+ "std": [
126
+ 0.3115188437483966
127
+ ],
128
+ "q01": [
129
+ 0.05372234335522662
130
+ ],
131
+ "q99": [
132
+ 1.42285219032216
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-30000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13a6a938b8f5f24748d3fbfff60561c9389ac355de54d0231019c645b6341064
3
+ size 4999367032
checkpoint-30000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ed710082d6cd25e97a7c33b9dbc05deee76524f7dc2775dcfe6a3bd4a7665fd
3
+ size 2586705312
checkpoint-30000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3855fe2ba19da19528472666288bd51109138a81e0869dfb79fe20e7df65189b
3
+ size 10210298785
checkpoint-30000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd6de0cbe73d97c27686edc8ff71e4f68c58338e2dacdcd87559b9b41cb49c61
3
+ size 14244
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe8c620e9a7808e10f2bc64fa81fe5b1efb6dcc2acb3596ed32f58119a5368c1
3
+ size 1064
checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
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
+ 1.6115039587020874,
8
+ 3.1076648235321045,
9
+ 0.2685627043247223,
10
+ 1.6672509908676147,
11
+ 1.969820261001587
12
+ ],
13
+ "min": [
14
+ -1.908600926399231,
15
+ -0.17690595984458923,
16
+ -2.865432024002075,
17
+ -2.654787063598633,
18
+ -1.9672902822494507
19
+ ],
20
+ "mean": [
21
+ -0.00959783991910409,
22
+ 1.499808118712209,
23
+ -1.2723766745625318,
24
+ 0.6897374995980092,
25
+ 0.012068234835772905
26
+ ],
27
+ "std": [
28
+ 0.8033180848336506,
29
+ 0.6881857246367352,
30
+ 0.5993451999991707,
31
+ 0.47767480027308595,
32
+ 0.8412626329126346
33
+ ],
34
+ "q01": [
35
+ -1.4122188767992025,
36
+ 0.04034731964515659,
37
+ -2.793325455606157,
38
+ -0.9005685478050315,
39
+ -1.4498776078523188
40
+ ],
41
+ "q99": [
42
+ 1.4172475041389343,
43
+ 2.8561244349134496,
44
+ -0.21244733821766565,
45
+ 1.2000015139915712,
46
+ 1.4504311649198878
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 1.9673068523406982
52
+ ],
53
+ "min": [
54
+ -0.03646281734108925
55
+ ],
56
+ "mean": [
57
+ 0.4539852668266194
58
+ ],
59
+ "std": [
60
+ 0.31508159664095936
61
+ ],
62
+ "q01": [
63
+ 0.05382254484269338
64
+ ],
65
+ "q99": [
66
+ 1.4516974584290336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 1.6115039587020874,
74
+ 3.1076648235321045,
75
+ 0.24772115051746368,
76
+ 1.6672509908676147,
77
+ 1.9685454368591309
78
+ ],
79
+ "min": [
80
+ -1.908600926399231,
81
+ -0.11258061230182648,
82
+ -2.865432024002075,
83
+ -2.65680193901062,
84
+ -1.9372831583023071
85
+ ],
86
+ "mean": [
87
+ -0.00968967524832796,
88
+ 1.506859118246074,
89
+ -1.2779861339022518,
90
+ 0.6942818922168711,
91
+ 0.012125308062864543
92
+ ],
93
+ "std": [
94
+ 0.8053480669448868,
95
+ 0.6855857436916613,
96
+ 0.5989715373901173,
97
+ 0.47624700980912044,
98
+ 0.8405142018342715
99
+ ],
100
+ "q01": [
101
+ -1.4130165935286558,
102
+ 0.044443692726112066,
103
+ -2.7953697018785206,
104
+ -0.8843341536846144,
105
+ -1.4473891171261197
106
+ ],
107
+ "q99": [
108
+ 1.417625653380196,
109
+ 2.8571220874299215,
110
+ -0.2219349648594896,
111
+ 1.2000015139915712,
112
+ 1.4479686097661741
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 1.937300443649292
118
+ ],
119
+ "min": [
120
+ -0.03744976595044136
121
+ ],
122
+ "mean": [
123
+ 0.4494256477809244
124
+ ],
125
+ "std": [
126
+ 0.3115188437483966
127
+ ],
128
+ "q01": [
129
+ 0.05372234335522662
130
+ ],
131
+ "q99": [
132
+ 1.42285219032216
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
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13a6a938b8f5f24748d3fbfff60561c9389ac355de54d0231019c645b6341064
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ed710082d6cd25e97a7c33b9dbc05deee76524f7dc2775dcfe6a3bd4a7665fd
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:ee9957990abc72bf2c56e92c5a91b5dc4859f30a7c3df2b5b6c4c786f5f97c51
3
+ size 5368