add gr00t15 lift cube
Browse files- checkpoint-9000/config.json +64 -0
- checkpoint-9000/experiment_cfg/metadata.json +195 -0
- checkpoint-9000/model-00001-of-00002.safetensors +3 -0
- checkpoint-9000/model-00002-of-00002.safetensors +3 -0
- checkpoint-9000/model.safetensors.index.json +0 -0
- checkpoint-9000/optimizer.pt +3 -0
- checkpoint-9000/rng_state.pth +3 -0
- checkpoint-9000/scheduler.pt +3 -0
- checkpoint-9000/trainer_state.json +0 -0
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 |
+
0.7608810067176819,
|
| 8 |
+
1.085809588432312,
|
| 9 |
+
1.528531789779663,
|
| 10 |
+
1.657511830329895,
|
| 11 |
+
1.5684120655059814
|
| 12 |
+
],
|
| 13 |
+
"min": [
|
| 14 |
+
-0.718605637550354,
|
| 15 |
+
-1.6991723775863647,
|
| 16 |
+
-1.486725926399231,
|
| 17 |
+
0.21916155517101288,
|
| 18 |
+
-0.9440526962280273
|
| 19 |
+
],
|
| 20 |
+
"mean": [
|
| 21 |
+
0.023931235074996948,
|
| 22 |
+
-0.29475539922714233,
|
| 23 |
+
0.3620274066925049,
|
| 24 |
+
1.1636991500854492,
|
| 25 |
+
0.09153899550437927
|
| 26 |
+
],
|
| 27 |
+
"std": [
|
| 28 |
+
0.24707143008708954,
|
| 29 |
+
0.6817215085029602,
|
| 30 |
+
0.4626528024673462,
|
| 31 |
+
0.23700019717216492,
|
| 32 |
+
0.39124080538749695
|
| 33 |
+
],
|
| 34 |
+
"q01": [
|
| 35 |
+
-0.5169710493087769,
|
| 36 |
+
-1.649198751449585,
|
| 37 |
+
-0.8802185940742493,
|
| 38 |
+
0.6561568593978881,
|
| 39 |
+
-0.7151143956184387
|
| 40 |
+
],
|
| 41 |
+
"q99": [
|
| 42 |
+
0.5810562515258794,
|
| 43 |
+
0.831950228214264,
|
| 44 |
+
1.2774355220794678,
|
| 45 |
+
1.6205930376052857,
|
| 46 |
+
1.1468248462677009
|
| 47 |
+
]
|
| 48 |
+
},
|
| 49 |
+
"gripper": {
|
| 50 |
+
"max": [
|
| 51 |
+
1.3606617450714111
|
| 52 |
+
],
|
| 53 |
+
"min": [
|
| 54 |
+
-0.17453302443027496
|
| 55 |
+
],
|
| 56 |
+
"mean": [
|
| 57 |
+
0.4436603784561157
|
| 58 |
+
],
|
| 59 |
+
"std": [
|
| 60 |
+
0.39972078800201416
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
-0.17453300952911377
|
| 64 |
+
],
|
| 65 |
+
"q99": [
|
| 66 |
+
1.345034146308899
|
| 67 |
+
]
|
| 68 |
+
}
|
| 69 |
+
},
|
| 70 |
+
"action": {
|
| 71 |
+
"single_arm": {
|
| 72 |
+
"max": [
|
| 73 |
+
0.7760240435600281,
|
| 74 |
+
1.090288519859314,
|
| 75 |
+
1.5558991432189941,
|
| 76 |
+
1.6323895454406738,
|
| 77 |
+
1.639390230178833
|
| 78 |
+
],
|
| 79 |
+
"min": [
|
| 80 |
+
-0.734919548034668,
|
| 81 |
+
-1.736653208732605,
|
| 82 |
+
-1.5576239824295044,
|
| 83 |
+
0.11909526586532593,
|
| 84 |
+
-0.9371041059494019
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.02507529780268669,
|
| 88 |
+
-0.3058025538921356,
|
| 89 |
+
0.3674618899822235,
|
| 90 |
+
1.176397442817688,
|
| 91 |
+
0.09076160192489624
|
| 92 |
+
],
|
| 93 |
+
"std": [
|
| 94 |
+
0.2511993646621704,
|
| 95 |
+
0.733898401260376,
|
| 96 |
+
0.47884073853492737,
|
| 97 |
+
0.24278493225574493,
|
| 98 |
+
0.3971458971500397
|
| 99 |
+
],
|
| 100 |
+
"q01": [
|
| 101 |
+
-0.529397189617157,
|
| 102 |
+
-1.704841136932373,
|
| 103 |
+
-0.8966632866859436,
|
| 104 |
+
0.6568068265914917,
|
| 105 |
+
-0.729019284248352
|
| 106 |
+
],
|
| 107 |
+
"q99": [
|
| 108 |
+
0.5988495945930481,
|
| 109 |
+
0.8777256608009338,
|
| 110 |
+
1.3220124244689941,
|
| 111 |
+
1.6224055290222168,
|
| 112 |
+
1.1264034509658813
|
| 113 |
+
]
|
| 114 |
+
},
|
| 115 |
+
"gripper": {
|
| 116 |
+
"max": [
|
| 117 |
+
1.4851993322372437
|
| 118 |
+
],
|
| 119 |
+
"min": [
|
| 120 |
+
-0.34853291511535645
|
| 121 |
+
],
|
| 122 |
+
"mean": [
|
| 123 |
+
0.27568525075912476
|
| 124 |
+
],
|
| 125 |
+
"std": [
|
| 126 |
+
0.5549324154853821
|
| 127 |
+
],
|
| 128 |
+
"q01": [
|
| 129 |
+
-0.31658151745796204
|
| 130 |
+
],
|
| 131 |
+
"q99": [
|
| 132 |
+
1.3532261848449707
|
| 133 |
+
]
|
| 134 |
+
}
|
| 135 |
+
}
|
| 136 |
+
},
|
| 137 |
+
"modalities": {
|
| 138 |
+
"video": {
|
| 139 |
+
"front": {
|
| 140 |
+
"resolution": [
|
| 141 |
+
480,
|
| 142 |
+
480
|
| 143 |
+
],
|
| 144 |
+
"channels": 3,
|
| 145 |
+
"fps": 30.0
|
| 146 |
+
},
|
| 147 |
+
"wrist": {
|
| 148 |
+
"resolution": [
|
| 149 |
+
480,
|
| 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 sha256:0e65895a7b6c052e58b679bdd68bff54a68694a81f38e23d3e03a92059b3600b
|
| 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 sha256:539f0aeafc3183835ee176e46aac642fca795ea4482dea7a6a45c75936d67301
|
| 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/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e236c813d1b0432b94a8c5258b417bf65ea9f811438b8efe4332b6c9f66df88
|
| 3 |
+
size 4147424698
|
checkpoint-9000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1730afc15e18cbcf58c043591a87f29fd00eb674b14e1c8786b06524bbf12a78
|
| 3 |
+
size 14244
|
checkpoint-9000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1c26823aa5685497db4fe82a648513d12cb6c26d9be110887b197c10fe3a44ac
|
| 3 |
+
size 1064
|
checkpoint-9000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|