Hugging Face's logo

OPTML-Group
/
ResWAM-LIBERO

Robotics
PyTorch
vision-language-action
world-action-model
flow-matching
libero
libero-plus
Model card Files
xet
 Community
ResWAM-LIBERO
File size: 4,705 Bytes
9303a54
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
{
  "action": {
    "environment_dim": 7,
    "horizon": 8,
    "representation": "LIBERO 7-D delta EEF (xyz, rotation, gripper)"
  },
  "benchmark": "libero_plus",
  "data": {
    "action_codec": {
      "continuous": "dims 0-5 z-score, mean/std over all 273,465 real frames (float64, ddof 0)",
      "decode_clip": false,
      "encode_clamp_sigma": 12.0,
      "gripper_decode": ">= 0 -> +1, else -1",
      "gripper_target": "+-1 (raw > 0.5 -> +1 close, else -1 open)",
      "statistics": "codec_statistics.json",
      "version": "reswam-libero-delta-eef-zscore6-binary-gripper-pm1-v1"
    }
  },
  "dataset": {
    "demonstrations_only": true,
    "end_of_episode_padding": {
      "action_mask": "all ones (padded slots supervised)",
      "action_past_end": "dims 0-5 zero delta (delta-EEF neutral action); gripper holds the last real value",
      "expected_padded_action_slots": 47404,
      "window_starts": "every real frame s in [0, L-1]",
      "world_future_frame": "t+8 clamps to the last real frame"
    },
    "expected_episodes": 1693,
    "expected_frames": 273465,
    "expected_legal_windows": 259921,
    "expected_padded_windows": 273465,
    "expected_tasks": 40,
    "sampling": "uniform_every_frame_windows_end_padded_then_distributed_shuffle",
    "source_repo": "physical-intelligence/libero",
    "suites": [
      "libero_spatial",
      "libero_object",
      "libero_goal",
      "libero_10"
    ],
    "task_balanced": false
  },
  "evaluation": {
    "action_horizon": 8,
    "benchmark_variant": "libero_plus",
    "replan_interval": 8,
    "seed": 0,
    "suites": [
      "libero_spatial",
      "libero_object",
      "libero_goal",
      "libero_10"
    ],
    "total_episodes": 10030,
    "trials_per_task": 1
  },
  "flow": {
    "action_noise": "independent_gaussian",
    "action_tau": "independent_uniform",
    "inference": {
      "rounds": 10,
      "solver": "synchronous_jacobi_euler",
      "synchronous_commit": true
    },
    "primary_noise": "independent_gaussian",
    "world_tau": "independent_uniform"
  },
  "loss": {
    "action": 1.0,
    "primary_residual": 0.5
  },
  "model": {
    "architecture": "Qwen3_5ForConditionalGeneration",
    "backbone_parameter_count": 1755487104,
    "dino": {
      "architecture": "DINOv3-B/16",
      "blocks": 12,
      "combine": "arithmetic_mean",
      "frozen": true,
      "image_size": 256,
      "model_id": "facebook/dinov3-vitb16-pretrain-lvd1689m",
      "output_grid": [
        8,
        8
      ],
      "patch_size": 16,
      "remove_cls_and_register": true,
      "selected_blocks": [
        8,
        10
      ],
      "spatial_pool": "2x2",
      "tokens_per_view": 64
    },
    "dino_parameter_count": 85660416,
    "expert": {
      "ffn_dim": 4096,
      "heads": 16,
      "hidden_dim": 1024,
      "joint_depth": 12,
      "parameter_streams": [
        "action",
        "primary_world"
      ],
      "use_state_null_tokens": true
    },
    "expert_parameter_count": 493578759,
    "repo": "Alibaba-DAMO-Academy/RynnBrain1.1-2B",
    "system_parameter_count": 2334726279,
    "text_hidden_size": 2048,
    "text_layers": 16,
    "trainable_parameter_count": 2249065863,
    "vision_hidden_size": 1024,
    "vision_layers": 24,
    "vision_output_size": 2048
  },
  "optimizer": {
    "betas": [
      0.9,
      0.95
    ],
    "expert_min_lr": 1e-05,
    "expert_peak_lr": 0.0001,
    "vlm_min_lr": 1e-06,
    "vlm_peak_lr": 1e-05,
    "weight_decay": 1e-08
  },
  "schema_version": "reswam-libero-config-v1",
  "state": {
    "clamp_sigma": 10.0,
    "dropout": {
      "probability": 0.3,
      "replacement": "learned null tokens state_null [1,11,dim]",
      "unit": "per sample (all 11 tokens)"
    },
    "features": [
      "eef_x",
      "eef_y",
      "eef_z",
      "rot_c0_x",
      "rot_c0_y",
      "rot_c0_z",
      "rot_c1_x",
      "rot_c1_y",
      "rot_c1_z",
      "finger_left",
      "finger_right"
    ],
    "featurization": "reswam-libero-state11-xyz-rot6d-fingers2-v1",
    "normalization": "z-score over all 273,465 real frames (float64, ddof 0); clamp +-10 sigma",
    "raw_layout": "eef xyz, axis-angle(robot0_eef_quat), robot0_gripper_qpos x2",
    "statistics": "codec_statistics.json"
  },
  "training": {
    "fsdp": "FULL_SHARD",
    "global_batch": 256,
    "gradient_accumulation": 4,
    "per_gpu_microbatch": 32,
    "precision": "bfloat16",
    "seed": 20260830,
    "world_size": 2
  },
  "world": {
    "policy_context_views": [
      "primary",
      "wrist"
    ],
    "state_views": [
      "primary"
    ],
    "target": "LN(DINO_blocks8_10_mean_pool2x2(o_t+8))-LN(DINO_blocks8_10_mean_pool2x2(o_t))",
    "target_offset": 8
  }
}