Hugging Face's logo

nvidia
/
Cosmos-Policy-ALOHA-Planning-Model-Predict2-2B

cosmos-policy-planning
Model card Files
xet
 Community
2
Cosmos-Policy-ALOHA-Planning-Model-Predict2-2B
File size: 1,787 Bytes
160f84b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
  "model_type": "cosmos-policy-planning",
  "architecture": "diffusion-transformer",
  "base_model": "nvidia/Cosmos-Policy-ALOHA-Predict2-2B",
  "num_parameters": "2B",

  "input_spec": {
    "text": {
      "type": "string",
      "description": "Natural language task description"
    },
    "images": {
      "format": "RGB",
      "resolution": [224, 224],
      "views": ["top_down", "left_wrist", "right_wrist"]
    },
    "proprioception": {
      "dim": 14,
      "components": ["left_arm_joints", "right_arm_joints"],
      "joints_per_arm": 7
    },
    "actions": {
      "dim": 14,
      "horizon": 50,
      "description": "Candidate action sequence to evaluate"
    }
  },

  "output_spec": {
    "future_proprioception": {
      "dim": 14
    },
    "future_images": {
      "resolution": [224, 224],
      "views": 3
    },
    "value": {
      "dim": 1,
      "description": "Expected cumulative reward for action sequence"
    }
  },

  "diffusion_config": {
    "denoising_steps": 10,
    "sigma_min": 4.0,
    "sigma_max": 80.0
  },

  "planning_config": {
    "ensemble_world_model_queries": 3,
    "ensemble_value_queries": 5,
    "total_predictions_per_action": 15,
    "best_of_n_search": 8
  },

  "training": {
    "dataset": "ALOHA policy rollouts",
    "num_episodes": 648,
    "hardware": "8x H100",
    "batch_split": {
      "policy": 0.1,
      "world_model": 0.45,
      "value_function": 0.45
    }
  },

  "benchmark_results": {
    "put_candies_in_bowl": 0.60,
    "put_candy_in_ziploc_bag": 0.84,
    "average": 0.72,
    "improvement_over_base": 0.125
  },

  "inference": {
    "precision": "bf16",
    "latency_seconds": 4.9,
    "recommended_gpus": 8
  },

  "robot_platform": "ALOHA 2 (ViperX 300 S dual arms)",
  "control_frequency_hz": 25
}