Hugging Face's logo

huru33
/
pickup-boxs16-model

TensorBoard
Safetensors
gr00t_n1_5
Model card Files
xet
Metrics Community
huru33 commited on
Commit
bc133de
·
verified ·
1 Parent(s): 21f68db

Add files using upload-large-folder tool

Browse files
Files changed (50) 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-6000/config.json +64 -0
  11. checkpoint-6000/experiment_cfg/metadata.json +195 -0
  12. checkpoint-6000/model-00001-of-00002.safetensors +3 -0
  13. checkpoint-6000/model-00002-of-00002.safetensors +3 -0
  14. checkpoint-6000/model.safetensors.index.json +0 -0
  15. checkpoint-6000/optimizer.pt +3 -0
  16. checkpoint-6000/rng_state.pth +3 -0
  17. checkpoint-6000/scheduler.pt +3 -0
  18. checkpoint-6000/trainer_state.json +0 -0
  19. checkpoint-7000/config.json +64 -0
  20. checkpoint-7000/experiment_cfg/metadata.json +195 -0
  21. checkpoint-7000/model-00001-of-00002.safetensors +3 -0
  22. checkpoint-7000/model-00002-of-00002.safetensors +3 -0
  23. checkpoint-7000/model.safetensors.index.json +0 -0
  24. checkpoint-7000/rng_state.pth +3 -0
  25. checkpoint-7000/scheduler.pt +3 -0
  26. checkpoint-7000/trainer_state.json +0 -0
  27. checkpoint-8000/config.json +64 -0
  28. checkpoint-8000/experiment_cfg/metadata.json +195 -0
  29. checkpoint-8000/model-00001-of-00002.safetensors +3 -0
  30. checkpoint-8000/model-00002-of-00002.safetensors +3 -0
  31. checkpoint-8000/model.safetensors.index.json +0 -0
  32. checkpoint-8000/rng_state.pth +3 -0
  33. checkpoint-8000/scheduler.pt +3 -0
  34. checkpoint-8000/trainer_state.json +0 -0
  35. checkpoint-9000/config.json +64 -0
  36. checkpoint-9000/experiment_cfg/metadata.json +195 -0
  37. checkpoint-9000/model-00001-of-00002.safetensors +3 -0
  38. checkpoint-9000/model-00002-of-00002.safetensors +3 -0
  39. checkpoint-9000/model.safetensors.index.json +0 -0
  40. checkpoint-9000/rng_state.pth +3 -0
  41. checkpoint-9000/scheduler.pt +3 -0
  42. checkpoint-9000/trainer_state.json +0 -0
  43. config.json +64 -0
  44. experiment_cfg/metadata.json +195 -0
  45. model-00001-of-00002.safetensors +3 -0
  46. model-00002-of-00002.safetensors +3 -0
  47. model.safetensors.index.json +0 -0
  48. runs/Oct11_10-52-40_autodl-container-08a34fb751-80978039/events.out.tfevents.1760151163.autodl-container-08a34fb751-80978039.270193.0 +3 -0
  49. trainer_state.json +0 -0
  50. 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
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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 sha256:10dc4e125806514e3f82c57fc253ddaa11aa7ce58d50639101766255dff304ed
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 sha256:9daddbf951935dafe262b269c58226a92929e21e227582fb656735142a56d7e0
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 sha256:42c4a5970d19dc296ba544e71d536d33b38ff1870e4cd8ed1148f45a7fa8a8e9
3
+ size 8550719959
checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0298afc20cc72ae6d25eb2ddce0a37554323d69c2f6d003de306ce44ac3ee129
3
+ size 14645
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5028d488d4be5f20c2efa67f74447f402f4068511c82e3ac5f579bcb23ae1045
3
+ size 1465
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-6000/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-6000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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-6000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cce65dc330d482779a11cf1fe0796c78f965d0d1a876c9508eb2c287baf39e62
3
+ size 4999367032
checkpoint-6000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b4e73ee12f2d5847b23d7f170a64cf1b90e1522de9d5d872d072d00d2380c24
3
+ size 2586705312
checkpoint-6000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-6000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1564b3c3d4a72544c24932e2040fd2a623928995a9be20ba1c24111475a2acaa
3
+ size 8550719959
checkpoint-6000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8b706cafb52cb01ab1af7ade4157c794466fc9d94475b4acc45760810f7b1e0
3
+ size 14645
checkpoint-6000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bc8794e87a2f0ef4d469dfba980f0472f491c4ec8155a0864d4d936f98b1d40
3
+ size 1465
checkpoint-6000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-7000/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-7000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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-7000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88330906a5b509d235e9eaf5a606406f304438b320bab141c86f35822fd88fb6
3
+ size 4999367032
checkpoint-7000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85aa6a901d21a2dfaacf6d23becadd6e472300af98f5fa532c71a848cddc16cb
3
+ size 2586705312
checkpoint-7000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-7000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2b41e2757d885431978b361de5be69bcbe7636155e57c4e008abc2e6c9ecc0d
3
+ size 14645
checkpoint-7000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e877e3e674dbc48cbd9e109a8ee5cbe853b9f99d6b1c995c7f07780458e4567
3
+ size 1465
checkpoint-7000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-8000/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-8000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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-8000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7e5d5e3ea824d897639f2dacb746052496d6296160b8f848873ee6bfba5ff35
3
+ size 4999367032
checkpoint-8000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5db13b96257b834363139d1085e93a7b4c8ddcbd1fd9153b24b21cbfa64cbbfc
3
+ size 2586705312
checkpoint-8000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-8000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e48fd517071a869458c561b931f636b76a8da5e2a2044cb54d4e6144643769e7
3
+ size 14645
checkpoint-8000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c288f66085bd6b4f2a8e21462e99149d617360070c2119ea6ca9080d8550206
3
+ size 1465
checkpoint-8000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/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-9000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "single_arm": {
6
+ "max": [
7
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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-9000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de655c69f1b732c7584fefe5f47e4a2fa99a001a257aeaca1e326dee7f59e141
3
+ size 4999367032
checkpoint-9000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74fb0ed40e769d67b3374ebc91b3f3f8690d9e95eda0b66b729e61faa083e410
3
+ size 2586705312
checkpoint-9000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4220ebccdb68b452cdc0c6b5d5a7de40329612a46343d8b3230160dbdda31f2c
3
+ size 14645
checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26803a5937835496ccc84787a867339be1d60285f3844bc8c1c385aa50d89d2b
3
+ size 1465
checkpoint-9000/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
+ 43.97346115112305,
8
+ 70.44210815429688,
9
+ 98.55985260009766,
10
+ 98.9610366821289,
11
+ 58.28125
12
+ ],
13
+ "min": [
14
+ -25.396240234375,
15
+ -98.48420715332031,
16
+ -98.10980987548828,
17
+ -98.26839447021484,
18
+ -68.33333587646484
19
+ ],
20
+ "mean": [
21
+ 16.535076141357422,
22
+ -40.214942932128906,
23
+ 37.648155212402344,
24
+ 0.16998249292373657,
25
+ 0.3340470492839813
26
+ ],
27
+ "std": [
28
+ 13.129369735717773,
29
+ 63.69690704345703,
30
+ 64.97482299804688,
31
+ 85.84671020507812,
32
+ 16.200212478637695
33
+ ],
34
+ "q01": [
35
+ -13.748618125915527,
36
+ -98.4000015258789,
37
+ -84.33843231201172,
38
+ -98.18181610107422,
39
+ -53.17708206176758
40
+ ],
41
+ "q99": [
42
+ 39.62403106689453,
43
+ 62.652631759643555,
44
+ 98.4698486328125,
45
+ 98.70130157470703,
46
+ 40.390625
47
+ ]
48
+ },
49
+ "gripper": {
50
+ "max": [
51
+ 83.85175323486328
52
+ ],
53
+ "min": [
54
+ 1.588352084159851
55
+ ],
56
+ "mean": [
57
+ 11.162861824035645
58
+ ],
59
+ "std": [
60
+ 15.795912742614746
61
+ ],
62
+ "q01": [
63
+ 1.6545333862304688
64
+ ],
65
+ "q99": [
66
+ 61.317007064819336
67
+ ]
68
+ }
69
+ },
70
+ "action": {
71
+ "single_arm": {
72
+ "max": [
73
+ 44.34004592895508,
74
+ 69.5270004272461,
75
+ 100.0,
76
+ 99.9148178100586,
77
+ 58.818878173828125
78
+ ],
79
+ "min": [
80
+ -25.995525360107422,
81
+ -100.0,
82
+ -100.0,
83
+ -100.0,
84
+ -68.62641906738281
85
+ ],
86
+ "mean": [
87
+ 16.654430389404297,
88
+ -41.189964294433594,
89
+ 37.494991302490234,
90
+ -0.41649922728538513,
91
+ 0.34309351444244385
92
+ ],
93
+ "std": [
94
+ 13.177230834960938,
95
+ 63.16563415527344,
96
+ 66.31660461425781,
97
+ 86.57202911376953,
98
+ 16.224573135375977
99
+ ],
100
+ "q01": [
101
+ -13.64653205871582,
102
+ -100.0,
103
+ -86.69081115722656,
104
+ -99.48892974853516,
105
+ -53.17690658569336
106
+ ],
107
+ "q99": [
108
+ 39.552574157714844,
109
+ 61.322731018066406,
110
+ 100.0,
111
+ 99.31856536865234,
112
+ 40.73292922973633
113
+ ]
114
+ },
115
+ "gripper": {
116
+ "max": [
117
+ 84.70209503173828
118
+ ],
119
+ "min": [
120
+ 0.0
121
+ ],
122
+ "mean": [
123
+ 10.208662986755371
124
+ ],
125
+ "std": [
126
+ 16.52983283996582
127
+ ],
128
+ "q01": [
129
+ 0.6441223621368408
130
+ ],
131
+ "q99": [
132
+ 62.801931381225586
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 sha256:10dc4e125806514e3f82c57fc253ddaa11aa7ce58d50639101766255dff304ed
3
+ size 4999367032
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9daddbf951935dafe262b269c58226a92929e21e227582fb656735142a56d7e0
3
+ size 2586705312
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
runs/Oct11_10-52-40_autodl-container-08a34fb751-80978039/events.out.tfevents.1760151163.autodl-container-08a34fb751-80978039.270193.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e46f2e4f42b3466d3f4b65e30fa445262a5f1ede16e47e46a3c7bc368b34fb82
3
+ size 217395
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:ef2d3a36f9dec59ab91752ca212c81bc97c0fd8ba3332a04dd568397c72e80b7
3
+ size 5777