Upload folder using huggingface_hub
Browse files- config.json +131 -0
- model.safetensors +3 -0
- stats.json +446 -0
- train_config.json +333 -0
config.json
ADDED
|
@@ -0,0 +1,131 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "cubev2",
|
| 3 |
+
"n_obs_steps": 1,
|
| 4 |
+
"input_features": {
|
| 5 |
+
"observation.state": {
|
| 6 |
+
"type": "STATE",
|
| 7 |
+
"shape": [
|
| 8 |
+
32
|
| 9 |
+
]
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
+
"output_features": {
|
| 13 |
+
"action": {
|
| 14 |
+
"type": "ACTION",
|
| 15 |
+
"shape": [
|
| 16 |
+
32
|
| 17 |
+
]
|
| 18 |
+
}
|
| 19 |
+
},
|
| 20 |
+
"device": "cuda:0",
|
| 21 |
+
"use_amp": false,
|
| 22 |
+
"push_to_hub": false,
|
| 23 |
+
"repo_id": "lerobot_lab/WSA_Base",
|
| 24 |
+
"private": null,
|
| 25 |
+
"tags": null,
|
| 26 |
+
"license": null,
|
| 27 |
+
"pretrained_path": null,
|
| 28 |
+
"qwen3_vl_variant": "qwen3_vl_28l",
|
| 29 |
+
"action_expert_variant": "qwen3_28l",
|
| 30 |
+
"qwen3_vl_pretrained_path": "/data/jjhao/data/model/Qwen3-VL-2B-Instruct",
|
| 31 |
+
"dtype": "bfloat16",
|
| 32 |
+
"chunk_size": 50,
|
| 33 |
+
"n_action_steps": 50,
|
| 34 |
+
"max_state_dim": 32,
|
| 35 |
+
"max_action_dim": 32,
|
| 36 |
+
"mask_action_dim_padding_loss": false,
|
| 37 |
+
"action_loss_valid_dim": null,
|
| 38 |
+
"num_inference_steps": 10,
|
| 39 |
+
"time_sampling_beta_alpha": 1.5,
|
| 40 |
+
"time_sampling_beta_beta": 1.0,
|
| 41 |
+
"time_sampling_scale": 0.999,
|
| 42 |
+
"time_sampling_offset": 0.001,
|
| 43 |
+
"min_period": 0.004,
|
| 44 |
+
"max_period": 4.0,
|
| 45 |
+
"attention_mask_mode": "default",
|
| 46 |
+
"image_resolution": [
|
| 47 |
+
224,
|
| 48 |
+
224
|
| 49 |
+
],
|
| 50 |
+
"image_delta_indices": [
|
| 51 |
+
-15,
|
| 52 |
+
0,
|
| 53 |
+
15
|
| 54 |
+
],
|
| 55 |
+
"empty_cameras": 0,
|
| 56 |
+
"normalization_mapping": {
|
| 57 |
+
"VISUAL": "IDENTITY",
|
| 58 |
+
"STATE": "IDENTITY",
|
| 59 |
+
"ACTION": "IDENTITY"
|
| 60 |
+
},
|
| 61 |
+
"gradient_checkpointing": false,
|
| 62 |
+
"compile_model": false,
|
| 63 |
+
"compile_mode": "max-autotune",
|
| 64 |
+
"optimizer_lr": 5e-05,
|
| 65 |
+
"optimizer_betas": [
|
| 66 |
+
0.9,
|
| 67 |
+
0.95
|
| 68 |
+
],
|
| 69 |
+
"optimizer_eps": 1e-08,
|
| 70 |
+
"optimizer_weight_decay": 0.01,
|
| 71 |
+
"optimizer_grad_clip_norm": 1.0,
|
| 72 |
+
"scheduler_warmup_steps": 1000,
|
| 73 |
+
"scheduler_decay_steps": 150000,
|
| 74 |
+
"scheduler_decay_lr": 5e-06,
|
| 75 |
+
"tokenizer_max_length": 48,
|
| 76 |
+
"freeze_vision_encoder": false,
|
| 77 |
+
"train_expert_only": false,
|
| 78 |
+
"train_vlm_only": false,
|
| 79 |
+
"lora_modules": [],
|
| 80 |
+
"lora_unselected_mode": "full",
|
| 81 |
+
"lora_targets": [
|
| 82 |
+
"attn",
|
| 83 |
+
"ffn"
|
| 84 |
+
],
|
| 85 |
+
"lora_rank": 16,
|
| 86 |
+
"lora_alpha": 32.0,
|
| 87 |
+
"lora_rank_und": null,
|
| 88 |
+
"lora_alpha_und": null,
|
| 89 |
+
"lora_rank_gen": null,
|
| 90 |
+
"lora_alpha_gen": null,
|
| 91 |
+
"lora_rank_act": null,
|
| 92 |
+
"lora_alpha_act": null,
|
| 93 |
+
"lora_dropout": 0.0,
|
| 94 |
+
"scale_factor": 8,
|
| 95 |
+
"lambda_gen": 0.01,
|
| 96 |
+
"cosmos_tokenizer_path_or_name": "/data/jjhao/data/model/Cosmos-Tokenizer-CI8x8",
|
| 97 |
+
"enable_3d_queries": true,
|
| 98 |
+
"num_3d_query_tokens": 432,
|
| 99 |
+
"da3_alignment_mode": "query_decoder",
|
| 100 |
+
"da3_query_resampler_layers": 1,
|
| 101 |
+
"da3_query_resampler_ff_mult": 1,
|
| 102 |
+
"query_layer_indices": [
|
| 103 |
+
13,
|
| 104 |
+
19,
|
| 105 |
+
23,
|
| 106 |
+
27
|
| 107 |
+
],
|
| 108 |
+
"da3_variant": "large",
|
| 109 |
+
"da3_teacher_layers": [
|
| 110 |
+
11,
|
| 111 |
+
15,
|
| 112 |
+
19,
|
| 113 |
+
23
|
| 114 |
+
],
|
| 115 |
+
"da3_query_dim": 2048,
|
| 116 |
+
"da3_tokens_per_view": 1296,
|
| 117 |
+
"da3_num_views": 3,
|
| 118 |
+
"lambda_3d": 0.01,
|
| 119 |
+
"da3_model_path_or_name": "/data/jjhao/data/model/DA3-LARGE-1.1",
|
| 120 |
+
"da3_model_name": null,
|
| 121 |
+
"da3_code_root": "/data/jjhao/WSA_Dev/third_party/Depth-Anything-3",
|
| 122 |
+
"da3_teacher_process_res": 504,
|
| 123 |
+
"da3_layer_weights": [
|
| 124 |
+
1.0,
|
| 125 |
+
1.2,
|
| 126 |
+
1.4,
|
| 127 |
+
1.6
|
| 128 |
+
],
|
| 129 |
+
"future_query_init_std": 0.02,
|
| 130 |
+
"log_da3_teacher_timing": true
|
| 131 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f420dc17aaedb91b3a37ce7ba164dc8a4309b50e3f5d038ed35406887ac5164f
|
| 3 |
+
size 6608313224
|
stats.json
ADDED
|
@@ -0,0 +1,446 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"aloha": {
|
| 3 |
+
"observation.state": {
|
| 4 |
+
"min": [
|
| 5 |
+
-7.552278995513916,
|
| 6 |
+
-0.36354875564575195,
|
| 7 |
+
-0.18577136099338531,
|
| 8 |
+
-1.956291913986206,
|
| 9 |
+
-1.6801013946533203,
|
| 10 |
+
-7.678869724273682,
|
| 11 |
+
0.0,
|
| 12 |
+
-8.539926528930664,
|
| 13 |
+
-0.6294453740119934,
|
| 14 |
+
-0.07775841653347015,
|
| 15 |
+
-2.1328067779541016,
|
| 16 |
+
-2.1285502910614014,
|
| 17 |
+
-8.228925704956055,
|
| 18 |
+
0.0
|
| 19 |
+
],
|
| 20 |
+
"max": [
|
| 21 |
+
6.45659065246582,
|
| 22 |
+
4.179152488708496,
|
| 23 |
+
5.346591472625732,
|
| 24 |
+
1.7942548990249634,
|
| 25 |
+
1.8604620695114136,
|
| 26 |
+
5.43813943862915,
|
| 27 |
+
1.0,
|
| 28 |
+
7.142920017242432,
|
| 29 |
+
4.157068729400635,
|
| 30 |
+
5.672450065612793,
|
| 31 |
+
1.95806884765625,
|
| 32 |
+
1.5663840770721436,
|
| 33 |
+
5.278968811035156,
|
| 34 |
+
1.0
|
| 35 |
+
],
|
| 36 |
+
"mean": [
|
| 37 |
+
-0.23838327825069427,
|
| 38 |
+
1.129874348640442,
|
| 39 |
+
0.8068663477897644,
|
| 40 |
+
-0.3102293312549591,
|
| 41 |
+
0.05489448457956314,
|
| 42 |
+
-0.03798479959368706,
|
| 43 |
+
0.6655772924423218,
|
| 44 |
+
0.21665608882904053,
|
| 45 |
+
1.0790668725967407,
|
| 46 |
+
0.7875725626945496,
|
| 47 |
+
-0.3211246132850647,
|
| 48 |
+
-0.017505262047052383,
|
| 49 |
+
0.03539634123444557,
|
| 50 |
+
0.6794788837432861
|
| 51 |
+
],
|
| 52 |
+
"std": [
|
| 53 |
+
0.4151541292667389,
|
| 54 |
+
1.0045647621154785,
|
| 55 |
+
0.7768041491508484,
|
| 56 |
+
0.673305869102478,
|
| 57 |
+
0.27081969380378723,
|
| 58 |
+
0.6197779774665833,
|
| 59 |
+
0.4546463191509247,
|
| 60 |
+
0.3492097854614258,
|
| 61 |
+
1.024953842163086,
|
| 62 |
+
0.7943670153617859,
|
| 63 |
+
0.6847946047782898,
|
| 64 |
+
0.2443752884864807,
|
| 65 |
+
0.6142938733100891,
|
| 66 |
+
0.44977256655693054
|
| 67 |
+
],
|
| 68 |
+
"count": [
|
| 69 |
+
6075103
|
| 70 |
+
]
|
| 71 |
+
},
|
| 72 |
+
"timestamp": {
|
| 73 |
+
"min": [
|
| 74 |
+
0.0
|
| 75 |
+
],
|
| 76 |
+
"max": [
|
| 77 |
+
4.533333333333333
|
| 78 |
+
],
|
| 79 |
+
"mean": [
|
| 80 |
+
1.6372892807675414
|
| 81 |
+
],
|
| 82 |
+
"std": [
|
| 83 |
+
0.9676290760418356
|
| 84 |
+
],
|
| 85 |
+
"count": [
|
| 86 |
+
49335
|
| 87 |
+
]
|
| 88 |
+
},
|
| 89 |
+
"action": {
|
| 90 |
+
"min": [
|
| 91 |
+
-7.552278995513916,
|
| 92 |
+
-0.36354875564575195,
|
| 93 |
+
-0.18577136099338531,
|
| 94 |
+
-1.956291913986206,
|
| 95 |
+
-1.6801013946533203,
|
| 96 |
+
-7.678869724273682,
|
| 97 |
+
0.0,
|
| 98 |
+
-8.539926528930664,
|
| 99 |
+
-0.6294453740119934,
|
| 100 |
+
-0.07775841653347015,
|
| 101 |
+
-2.1328067779541016,
|
| 102 |
+
-2.1285502910614014,
|
| 103 |
+
-8.228925704956055,
|
| 104 |
+
0.0
|
| 105 |
+
],
|
| 106 |
+
"max": [
|
| 107 |
+
6.45659065246582,
|
| 108 |
+
4.179152488708496,
|
| 109 |
+
5.346591472625732,
|
| 110 |
+
1.7942548990249634,
|
| 111 |
+
1.8604620695114136,
|
| 112 |
+
5.43813943862915,
|
| 113 |
+
1.0,
|
| 114 |
+
7.142920017242432,
|
| 115 |
+
4.157068729400635,
|
| 116 |
+
5.672450065612793,
|
| 117 |
+
1.95806884765625,
|
| 118 |
+
1.5663840770721436,
|
| 119 |
+
5.278968811035156,
|
| 120 |
+
1.0
|
| 121 |
+
],
|
| 122 |
+
"mean": [
|
| 123 |
+
-0.23945418000221252,
|
| 124 |
+
1.1346182823181152,
|
| 125 |
+
0.8103252649307251,
|
| 126 |
+
-0.31154367327690125,
|
| 127 |
+
0.055183593183755875,
|
| 128 |
+
-0.03852220997214317,
|
| 129 |
+
0.6643287539482117,
|
| 130 |
+
0.2180495411157608,
|
| 131 |
+
1.0846160650253296,
|
| 132 |
+
0.791621744632721,
|
| 133 |
+
-0.32262614369392395,
|
| 134 |
+
-0.017654698342084885,
|
| 135 |
+
0.03604534640908241,
|
| 136 |
+
0.6783739924430847
|
| 137 |
+
],
|
| 138 |
+
"std": [
|
| 139 |
+
0.41571953892707825,
|
| 140 |
+
1.0039602518081665,
|
| 141 |
+
0.7768447995185852,
|
| 142 |
+
0.6748389005661011,
|
| 143 |
+
0.27154842019081116,
|
| 144 |
+
0.6218780875205994,
|
| 145 |
+
0.455098956823349,
|
| 146 |
+
0.349867582321167,
|
| 147 |
+
1.024841547012329,
|
| 148 |
+
0.7946923971176147,
|
| 149 |
+
0.6867099404335022,
|
| 150 |
+
0.24531899392604828,
|
| 151 |
+
0.6170235872268677,
|
| 152 |
+
0.4502118229866028
|
| 153 |
+
],
|
| 154 |
+
"count": [
|
| 155 |
+
6075103
|
| 156 |
+
]
|
| 157 |
+
},
|
| 158 |
+
"frame_index": {
|
| 159 |
+
"min": [
|
| 160 |
+
0
|
| 161 |
+
],
|
| 162 |
+
"max": [
|
| 163 |
+
136
|
| 164 |
+
],
|
| 165 |
+
"mean": [
|
| 166 |
+
49.118678423026246
|
| 167 |
+
],
|
| 168 |
+
"std": [
|
| 169 |
+
29.02887228125506
|
| 170 |
+
],
|
| 171 |
+
"count": [
|
| 172 |
+
49335
|
| 173 |
+
]
|
| 174 |
+
},
|
| 175 |
+
"task_index": {
|
| 176 |
+
"min": [
|
| 177 |
+
0
|
| 178 |
+
],
|
| 179 |
+
"max": [
|
| 180 |
+
488
|
| 181 |
+
],
|
| 182 |
+
"mean": [
|
| 183 |
+
241.3758791932705
|
| 184 |
+
],
|
| 185 |
+
"std": [
|
| 186 |
+
141.15870396113672
|
| 187 |
+
],
|
| 188 |
+
"count": [
|
| 189 |
+
49335
|
| 190 |
+
]
|
| 191 |
+
},
|
| 192 |
+
"episode_index": {
|
| 193 |
+
"min": [
|
| 194 |
+
0
|
| 195 |
+
],
|
| 196 |
+
"max": [
|
| 197 |
+
499
|
| 198 |
+
],
|
| 199 |
+
"mean": [
|
| 200 |
+
249.582101956015
|
| 201 |
+
],
|
| 202 |
+
"std": [
|
| 203 |
+
144.09172466866306
|
| 204 |
+
],
|
| 205 |
+
"count": [
|
| 206 |
+
49335
|
| 207 |
+
]
|
| 208 |
+
},
|
| 209 |
+
"observation.images.cam_high": {
|
| 210 |
+
"min": [
|
| 211 |
+
[
|
| 212 |
+
[
|
| 213 |
+
0.0
|
| 214 |
+
]
|
| 215 |
+
],
|
| 216 |
+
[
|
| 217 |
+
[
|
| 218 |
+
0.0
|
| 219 |
+
]
|
| 220 |
+
],
|
| 221 |
+
[
|
| 222 |
+
[
|
| 223 |
+
0.0
|
| 224 |
+
]
|
| 225 |
+
]
|
| 226 |
+
],
|
| 227 |
+
"max": [
|
| 228 |
+
[
|
| 229 |
+
[
|
| 230 |
+
1.0
|
| 231 |
+
]
|
| 232 |
+
],
|
| 233 |
+
[
|
| 234 |
+
[
|
| 235 |
+
1.0
|
| 236 |
+
]
|
| 237 |
+
],
|
| 238 |
+
[
|
| 239 |
+
[
|
| 240 |
+
1.0
|
| 241 |
+
]
|
| 242 |
+
]
|
| 243 |
+
],
|
| 244 |
+
"mean": [
|
| 245 |
+
[
|
| 246 |
+
[
|
| 247 |
+
0.48500001430511475
|
| 248 |
+
]
|
| 249 |
+
],
|
| 250 |
+
[
|
| 251 |
+
[
|
| 252 |
+
0.4560000002384186
|
| 253 |
+
]
|
| 254 |
+
],
|
| 255 |
+
[
|
| 256 |
+
[
|
| 257 |
+
0.4059999883174896
|
| 258 |
+
]
|
| 259 |
+
]
|
| 260 |
+
],
|
| 261 |
+
"std": [
|
| 262 |
+
[
|
| 263 |
+
[
|
| 264 |
+
0.2290000021457672
|
| 265 |
+
]
|
| 266 |
+
],
|
| 267 |
+
[
|
| 268 |
+
[
|
| 269 |
+
0.2240000069141388
|
| 270 |
+
]
|
| 271 |
+
],
|
| 272 |
+
[
|
| 273 |
+
[
|
| 274 |
+
0.22499999403953552
|
| 275 |
+
]
|
| 276 |
+
]
|
| 277 |
+
],
|
| 278 |
+
"count": [
|
| 279 |
+
48318
|
| 280 |
+
]
|
| 281 |
+
},
|
| 282 |
+
"observation.images.cam_left_wrist": {
|
| 283 |
+
"min": [
|
| 284 |
+
[
|
| 285 |
+
[
|
| 286 |
+
0.0
|
| 287 |
+
]
|
| 288 |
+
],
|
| 289 |
+
[
|
| 290 |
+
[
|
| 291 |
+
0.0
|
| 292 |
+
]
|
| 293 |
+
],
|
| 294 |
+
[
|
| 295 |
+
[
|
| 296 |
+
0.0
|
| 297 |
+
]
|
| 298 |
+
]
|
| 299 |
+
],
|
| 300 |
+
"max": [
|
| 301 |
+
[
|
| 302 |
+
[
|
| 303 |
+
1.0
|
| 304 |
+
]
|
| 305 |
+
],
|
| 306 |
+
[
|
| 307 |
+
[
|
| 308 |
+
1.0
|
| 309 |
+
]
|
| 310 |
+
],
|
| 311 |
+
[
|
| 312 |
+
[
|
| 313 |
+
1.0
|
| 314 |
+
]
|
| 315 |
+
]
|
| 316 |
+
],
|
| 317 |
+
"mean": [
|
| 318 |
+
[
|
| 319 |
+
[
|
| 320 |
+
0.48500001430511475
|
| 321 |
+
]
|
| 322 |
+
],
|
| 323 |
+
[
|
| 324 |
+
[
|
| 325 |
+
0.4560000002384186
|
| 326 |
+
]
|
| 327 |
+
],
|
| 328 |
+
[
|
| 329 |
+
[
|
| 330 |
+
0.4059999883174896
|
| 331 |
+
]
|
| 332 |
+
]
|
| 333 |
+
],
|
| 334 |
+
"std": [
|
| 335 |
+
[
|
| 336 |
+
[
|
| 337 |
+
0.2290000021457672
|
| 338 |
+
]
|
| 339 |
+
],
|
| 340 |
+
[
|
| 341 |
+
[
|
| 342 |
+
0.2240000069141388
|
| 343 |
+
]
|
| 344 |
+
],
|
| 345 |
+
[
|
| 346 |
+
[
|
| 347 |
+
0.22499999403953552
|
| 348 |
+
]
|
| 349 |
+
]
|
| 350 |
+
],
|
| 351 |
+
"count": [
|
| 352 |
+
48318
|
| 353 |
+
]
|
| 354 |
+
},
|
| 355 |
+
"observation.images.cam_right_wrist": {
|
| 356 |
+
"min": [
|
| 357 |
+
[
|
| 358 |
+
[
|
| 359 |
+
0.0
|
| 360 |
+
]
|
| 361 |
+
],
|
| 362 |
+
[
|
| 363 |
+
[
|
| 364 |
+
0.0
|
| 365 |
+
]
|
| 366 |
+
],
|
| 367 |
+
[
|
| 368 |
+
[
|
| 369 |
+
0.0
|
| 370 |
+
]
|
| 371 |
+
]
|
| 372 |
+
],
|
| 373 |
+
"max": [
|
| 374 |
+
[
|
| 375 |
+
[
|
| 376 |
+
1.0
|
| 377 |
+
]
|
| 378 |
+
],
|
| 379 |
+
[
|
| 380 |
+
[
|
| 381 |
+
1.0
|
| 382 |
+
]
|
| 383 |
+
],
|
| 384 |
+
[
|
| 385 |
+
[
|
| 386 |
+
1.0
|
| 387 |
+
]
|
| 388 |
+
]
|
| 389 |
+
],
|
| 390 |
+
"mean": [
|
| 391 |
+
[
|
| 392 |
+
[
|
| 393 |
+
0.48500001430511475
|
| 394 |
+
]
|
| 395 |
+
],
|
| 396 |
+
[
|
| 397 |
+
[
|
| 398 |
+
0.4560000002384186
|
| 399 |
+
]
|
| 400 |
+
],
|
| 401 |
+
[
|
| 402 |
+
[
|
| 403 |
+
0.4059999883174896
|
| 404 |
+
]
|
| 405 |
+
]
|
| 406 |
+
],
|
| 407 |
+
"std": [
|
| 408 |
+
[
|
| 409 |
+
[
|
| 410 |
+
0.2290000021457672
|
| 411 |
+
]
|
| 412 |
+
],
|
| 413 |
+
[
|
| 414 |
+
[
|
| 415 |
+
0.2240000069141388
|
| 416 |
+
]
|
| 417 |
+
],
|
| 418 |
+
[
|
| 419 |
+
[
|
| 420 |
+
0.22499999403953552
|
| 421 |
+
]
|
| 422 |
+
]
|
| 423 |
+
],
|
| 424 |
+
"count": [
|
| 425 |
+
48318
|
| 426 |
+
]
|
| 427 |
+
},
|
| 428 |
+
"index": {
|
| 429 |
+
"min": [
|
| 430 |
+
0
|
| 431 |
+
],
|
| 432 |
+
"max": [
|
| 433 |
+
49334
|
| 434 |
+
],
|
| 435 |
+
"mean": [
|
| 436 |
+
24667.0
|
| 437 |
+
],
|
| 438 |
+
"std": [
|
| 439 |
+
14241.78776230943
|
| 440 |
+
],
|
| 441 |
+
"count": [
|
| 442 |
+
49335
|
| 443 |
+
]
|
| 444 |
+
}
|
| 445 |
+
}
|
| 446 |
+
}
|
train_config.json
ADDED
|
@@ -0,0 +1,333 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": {
|
| 3 |
+
"type": "cubev2",
|
| 4 |
+
"repo_id": "multidata_from_file",
|
| 5 |
+
"repo_id_file": "/data/jjhao/WSA_Dev/outputs/WSA_Base/robotwin/_repo_id_files/WSA-Base-robotwin-abs-scratch-default-train-2026_08_08_13_01_13.txt",
|
| 6 |
+
"root": null,
|
| 7 |
+
"episodes": null,
|
| 8 |
+
"image_transforms": {
|
| 9 |
+
"enable": false,
|
| 10 |
+
"preset": null,
|
| 11 |
+
"max_num_transforms": 3,
|
| 12 |
+
"random_order": false,
|
| 13 |
+
"tfs": {
|
| 14 |
+
"brightness": {
|
| 15 |
+
"weight": 1.0,
|
| 16 |
+
"type": "ColorJitter",
|
| 17 |
+
"kwargs": {
|
| 18 |
+
"brightness": [
|
| 19 |
+
0.8,
|
| 20 |
+
1.2
|
| 21 |
+
]
|
| 22 |
+
}
|
| 23 |
+
},
|
| 24 |
+
"contrast": {
|
| 25 |
+
"weight": 1.0,
|
| 26 |
+
"type": "ColorJitter",
|
| 27 |
+
"kwargs": {
|
| 28 |
+
"contrast": [
|
| 29 |
+
0.8,
|
| 30 |
+
1.2
|
| 31 |
+
]
|
| 32 |
+
}
|
| 33 |
+
},
|
| 34 |
+
"saturation": {
|
| 35 |
+
"weight": 1.0,
|
| 36 |
+
"type": "ColorJitter",
|
| 37 |
+
"kwargs": {
|
| 38 |
+
"saturation": [
|
| 39 |
+
0.5,
|
| 40 |
+
1.5
|
| 41 |
+
]
|
| 42 |
+
}
|
| 43 |
+
},
|
| 44 |
+
"hue": {
|
| 45 |
+
"weight": 1.0,
|
| 46 |
+
"type": "ColorJitter",
|
| 47 |
+
"kwargs": {
|
| 48 |
+
"hue": [
|
| 49 |
+
-0.05,
|
| 50 |
+
0.05
|
| 51 |
+
]
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"sharpness": {
|
| 55 |
+
"weight": 1.0,
|
| 56 |
+
"type": "SharpnessJitter",
|
| 57 |
+
"kwargs": {
|
| 58 |
+
"sharpness": [
|
| 59 |
+
0.5,
|
| 60 |
+
1.5
|
| 61 |
+
]
|
| 62 |
+
}
|
| 63 |
+
},
|
| 64 |
+
"affine": {
|
| 65 |
+
"weight": 1.0,
|
| 66 |
+
"type": "RandomAffine",
|
| 67 |
+
"kwargs": {
|
| 68 |
+
"degrees": [
|
| 69 |
+
-5.0,
|
| 70 |
+
5.0
|
| 71 |
+
],
|
| 72 |
+
"translate": [
|
| 73 |
+
0.05,
|
| 74 |
+
0.05
|
| 75 |
+
]
|
| 76 |
+
}
|
| 77 |
+
}
|
| 78 |
+
}
|
| 79 |
+
},
|
| 80 |
+
"revision": null,
|
| 81 |
+
"use_imagenet_stats": true,
|
| 82 |
+
"use_external_stats": true,
|
| 83 |
+
"external_stats_path": "/data/jjhao/huggingface/lerobot/stats/aloha/abs/robotwin_clean_randomized_all/stats.json",
|
| 84 |
+
"external_stats_root": "/data/jjhao/huggingface/lerobot/stats",
|
| 85 |
+
"weight_rules_path": null,
|
| 86 |
+
"video_backend": "pyav",
|
| 87 |
+
"streaming": false,
|
| 88 |
+
"dist_loading": false,
|
| 89 |
+
"buffer_size": 1024,
|
| 90 |
+
"action_mode": "abs",
|
| 91 |
+
"repack_transforms": {
|
| 92 |
+
"inputs": [],
|
| 93 |
+
"outputs": []
|
| 94 |
+
},
|
| 95 |
+
"data_transforms": {
|
| 96 |
+
"inputs": [
|
| 97 |
+
{
|
| 98 |
+
"type": "resize_with_pad",
|
| 99 |
+
"height": 224,
|
| 100 |
+
"width": 224,
|
| 101 |
+
"mode": "bilinear"
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"type": "remap_image_key",
|
| 105 |
+
"mapping": {}
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"type": "normalize",
|
| 109 |
+
"selected_keys": null,
|
| 110 |
+
"mode": "mean_std",
|
| 111 |
+
"norm_stats": {}
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"type": "compose_fields",
|
| 115 |
+
"mapping": {}
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"type": "pad_state_and_action",
|
| 119 |
+
"max_state_dim": 32,
|
| 120 |
+
"max_action_dim": 32
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"type": "cubev2_processor",
|
| 124 |
+
"pretrained_model_name_or_path": "/data/jjhao/data/model/Qwen3-VL-2B-Instruct",
|
| 125 |
+
"max_length": 48,
|
| 126 |
+
"task_key": "task",
|
| 127 |
+
"padding_side": "right",
|
| 128 |
+
"padding": "max_length",
|
| 129 |
+
"truncation": true,
|
| 130 |
+
"spatial_merge_size": 2,
|
| 131 |
+
"vision_start_token_id": 151652,
|
| 132 |
+
"vision_end_token_id": 151653,
|
| 133 |
+
"image_token_id": 151655,
|
| 134 |
+
"processor": null,
|
| 135 |
+
"_processor_source": null
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"type": "unify_cubev2_inputs"
|
| 139 |
+
}
|
| 140 |
+
],
|
| 141 |
+
"outputs": []
|
| 142 |
+
},
|
| 143 |
+
"model_transforms": {
|
| 144 |
+
"inputs": [],
|
| 145 |
+
"outputs": []
|
| 146 |
+
},
|
| 147 |
+
"height": 224,
|
| 148 |
+
"width": 224,
|
| 149 |
+
"max_state_dim": 32,
|
| 150 |
+
"max_action_dim": 32,
|
| 151 |
+
"qwen3_vl_processor_path": "/data/jjhao/data/model/Qwen3-VL-2B-Instruct"
|
| 152 |
+
},
|
| 153 |
+
"env": null,
|
| 154 |
+
"policy": {
|
| 155 |
+
"type": "cubev2",
|
| 156 |
+
"n_obs_steps": 1,
|
| 157 |
+
"input_features": {
|
| 158 |
+
"observation.state": {
|
| 159 |
+
"type": "STATE",
|
| 160 |
+
"shape": [
|
| 161 |
+
32
|
| 162 |
+
]
|
| 163 |
+
}
|
| 164 |
+
},
|
| 165 |
+
"output_features": {
|
| 166 |
+
"action": {
|
| 167 |
+
"type": "ACTION",
|
| 168 |
+
"shape": [
|
| 169 |
+
32
|
| 170 |
+
]
|
| 171 |
+
}
|
| 172 |
+
},
|
| 173 |
+
"device": "cuda:0",
|
| 174 |
+
"use_amp": false,
|
| 175 |
+
"push_to_hub": false,
|
| 176 |
+
"repo_id": "lerobot_lab/WSA_Base",
|
| 177 |
+
"private": null,
|
| 178 |
+
"tags": null,
|
| 179 |
+
"license": null,
|
| 180 |
+
"pretrained_path": null,
|
| 181 |
+
"qwen3_vl_variant": "qwen3_vl_28l",
|
| 182 |
+
"action_expert_variant": "qwen3_28l",
|
| 183 |
+
"qwen3_vl_pretrained_path": "/data/jjhao/data/model/Qwen3-VL-2B-Instruct",
|
| 184 |
+
"dtype": "bfloat16",
|
| 185 |
+
"chunk_size": 50,
|
| 186 |
+
"n_action_steps": 50,
|
| 187 |
+
"max_state_dim": 32,
|
| 188 |
+
"max_action_dim": 32,
|
| 189 |
+
"mask_action_dim_padding_loss": false,
|
| 190 |
+
"action_loss_valid_dim": null,
|
| 191 |
+
"num_inference_steps": 10,
|
| 192 |
+
"time_sampling_beta_alpha": 1.5,
|
| 193 |
+
"time_sampling_beta_beta": 1.0,
|
| 194 |
+
"time_sampling_scale": 0.999,
|
| 195 |
+
"time_sampling_offset": 0.001,
|
| 196 |
+
"min_period": 0.004,
|
| 197 |
+
"max_period": 4.0,
|
| 198 |
+
"attention_mask_mode": "default",
|
| 199 |
+
"image_resolution": [
|
| 200 |
+
224,
|
| 201 |
+
224
|
| 202 |
+
],
|
| 203 |
+
"image_delta_indices": [
|
| 204 |
+
-15,
|
| 205 |
+
0,
|
| 206 |
+
15
|
| 207 |
+
],
|
| 208 |
+
"empty_cameras": 0,
|
| 209 |
+
"normalization_mapping": {
|
| 210 |
+
"VISUAL": "IDENTITY",
|
| 211 |
+
"STATE": "IDENTITY",
|
| 212 |
+
"ACTION": "IDENTITY"
|
| 213 |
+
},
|
| 214 |
+
"gradient_checkpointing": false,
|
| 215 |
+
"compile_model": false,
|
| 216 |
+
"compile_mode": "max-autotune",
|
| 217 |
+
"optimizer_lr": 5e-05,
|
| 218 |
+
"optimizer_betas": [
|
| 219 |
+
0.9,
|
| 220 |
+
0.95
|
| 221 |
+
],
|
| 222 |
+
"optimizer_eps": 1e-08,
|
| 223 |
+
"optimizer_weight_decay": 0.01,
|
| 224 |
+
"optimizer_grad_clip_norm": 1.0,
|
| 225 |
+
"scheduler_warmup_steps": 1000,
|
| 226 |
+
"scheduler_decay_steps": 150000,
|
| 227 |
+
"scheduler_decay_lr": 5e-06,
|
| 228 |
+
"tokenizer_max_length": 48,
|
| 229 |
+
"freeze_vision_encoder": false,
|
| 230 |
+
"train_expert_only": false,
|
| 231 |
+
"train_vlm_only": false,
|
| 232 |
+
"lora_modules": [],
|
| 233 |
+
"lora_unselected_mode": "full",
|
| 234 |
+
"lora_targets": [
|
| 235 |
+
"attn",
|
| 236 |
+
"ffn"
|
| 237 |
+
],
|
| 238 |
+
"lora_rank": 16,
|
| 239 |
+
"lora_alpha": 32.0,
|
| 240 |
+
"lora_rank_und": null,
|
| 241 |
+
"lora_alpha_und": null,
|
| 242 |
+
"lora_rank_gen": null,
|
| 243 |
+
"lora_alpha_gen": null,
|
| 244 |
+
"lora_rank_act": null,
|
| 245 |
+
"lora_alpha_act": null,
|
| 246 |
+
"lora_dropout": 0.0,
|
| 247 |
+
"scale_factor": 8,
|
| 248 |
+
"lambda_gen": 0.01,
|
| 249 |
+
"cosmos_tokenizer_path_or_name": "/data/jjhao/data/model/Cosmos-Tokenizer-CI8x8",
|
| 250 |
+
"enable_3d_queries": true,
|
| 251 |
+
"num_3d_query_tokens": 432,
|
| 252 |
+
"da3_alignment_mode": "query_decoder",
|
| 253 |
+
"da3_query_resampler_layers": 1,
|
| 254 |
+
"da3_query_resampler_ff_mult": 1,
|
| 255 |
+
"query_layer_indices": [
|
| 256 |
+
13,
|
| 257 |
+
19,
|
| 258 |
+
23,
|
| 259 |
+
27
|
| 260 |
+
],
|
| 261 |
+
"da3_variant": "large",
|
| 262 |
+
"da3_teacher_layers": [
|
| 263 |
+
11,
|
| 264 |
+
15,
|
| 265 |
+
19,
|
| 266 |
+
23
|
| 267 |
+
],
|
| 268 |
+
"da3_query_dim": 2048,
|
| 269 |
+
"da3_tokens_per_view": 1296,
|
| 270 |
+
"da3_num_views": 3,
|
| 271 |
+
"lambda_3d": 0.01,
|
| 272 |
+
"da3_model_path_or_name": "/data/jjhao/data/model/DA3-LARGE-1.1",
|
| 273 |
+
"da3_model_name": null,
|
| 274 |
+
"da3_code_root": "/data/jjhao/WSA_Dev/third_party/Depth-Anything-3",
|
| 275 |
+
"da3_teacher_process_res": 504,
|
| 276 |
+
"da3_layer_weights": [
|
| 277 |
+
1.0,
|
| 278 |
+
1.2,
|
| 279 |
+
1.4,
|
| 280 |
+
1.6
|
| 281 |
+
],
|
| 282 |
+
"future_query_init_std": 0.02,
|
| 283 |
+
"log_da3_teacher_timing": true
|
| 284 |
+
},
|
| 285 |
+
"output_dir": "/data/jjhao/WSA_Dev/outputs/WSA_Base/robotwin/WSA-Base-robotwin-abs-scratch-default-train-2026_08_08_13_01_13",
|
| 286 |
+
"job_name": "WSA-Base-robotwin-abs-scratch-default-train-2026_08_08_13_01_13",
|
| 287 |
+
"resume": false,
|
| 288 |
+
"seed": 42,
|
| 289 |
+
"num_workers": 12,
|
| 290 |
+
"batch_size": 8,
|
| 291 |
+
"gradient_accumulation_steps": 1,
|
| 292 |
+
"steps": 150000,
|
| 293 |
+
"eval_freq": 20000,
|
| 294 |
+
"eval_max_batches": 2,
|
| 295 |
+
"log_freq": 100,
|
| 296 |
+
"save_checkpoint": true,
|
| 297 |
+
"save_freq": 30000,
|
| 298 |
+
"use_policy_training_preset": true,
|
| 299 |
+
"optimizer": {
|
| 300 |
+
"type": "adamw",
|
| 301 |
+
"lr": 5e-05,
|
| 302 |
+
"weight_decay": 0.01,
|
| 303 |
+
"grad_clip_norm": 1.0,
|
| 304 |
+
"betas": [
|
| 305 |
+
0.9,
|
| 306 |
+
0.95
|
| 307 |
+
],
|
| 308 |
+
"eps": 1e-08
|
| 309 |
+
},
|
| 310 |
+
"scheduler": {
|
| 311 |
+
"type": "cosine_decay_with_warmup",
|
| 312 |
+
"num_warmup_steps": 1000,
|
| 313 |
+
"num_decay_steps": 150000,
|
| 314 |
+
"peak_lr": 5e-05,
|
| 315 |
+
"decay_lr": 5e-06
|
| 316 |
+
},
|
| 317 |
+
"eval": {
|
| 318 |
+
"n_episodes": 50,
|
| 319 |
+
"batch_size": 50,
|
| 320 |
+
"use_async_envs": false
|
| 321 |
+
},
|
| 322 |
+
"wandb": {
|
| 323 |
+
"enable": true,
|
| 324 |
+
"disable_artifact": true,
|
| 325 |
+
"project": "WSA_Base",
|
| 326 |
+
"entity": null,
|
| 327 |
+
"notes": null,
|
| 328 |
+
"run_id": "1tht6jso",
|
| 329 |
+
"mode": "offline"
|
| 330 |
+
},
|
| 331 |
+
"checkpoint_path": null,
|
| 332 |
+
"rename_map": {}
|
| 333 |
+
}
|