YAM/create_processor_configs.py at main · thomas0829/YAM · GitHub

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
#!/usr/bin/env python
"""
Create processor configuration files for pi05_pytorch model using dataset statistics.
"""

import json
from pathlib import Path

# Load dataset statistics
dataset_path = Path("datasets/pick_up_four_cubes_and_stack_them_in_the_middle-v3.0")
stats_path = dataset_path / "meta" / "stats.json"

with open(stats_path) as f:
    stats = json.load(f)

# Extract statistics for each feature
features = {}

# State feature (MIN_MAX normalization)
features["observation.state"] = {
    "type": "STATE",
    "shape": [6],
    "normalization": {
        "mode": "MIN_MAX",
        "min": stats["observation.state"]["min"],
        "max": stats["observation.state"]["max"]
    }
}

# Image features (MEAN_STD normalization)
for camera in ["top", "wrist"]:
    key = f"observation.images.{camera}"
    # Flatten nested lists to simple arrays
    mean = [stats[key]["mean"][i][0][0] for i in range(3)]
    std = [stats[key]["std"][i][0][0] for i in range(3)]

    features[key] = {
        "type": "VISUAL",
        "shape": [3, 224, 224],
        "normalization": {
            "mode": "MEAN_STD",
            "mean": mean,
            "std": std
        }
    }

# Action feature (MEAN_STD normalization)
features["action"] = {
    "type": "ACTION",
    "shape": [6],
    "normalization": {
        "mode": "MEAN_STD",
        "mean": stats["action"]["mean"],
        "std": stats["action"]["std"]
    }
}

# Create preprocessor configuration
preprocessor = {
    "name": "policy_preprocessor",
    "steps": [
        {
            "registry_name": "rename_observations_processor",
            "config": {
                "rename_map": {}
            }
        },
        {
            "registry_name": "to_batch_processor",
            "config": {}
        },
        {
            "registry_name": "normalizer_processor",
            "config": {
                "eps": 1e-08,
                "features": features,
                "norm_map": {
                    "VISUAL": "MEAN_STD",
                    "STATE": "MIN_MAX",
                    "ACTION": "MEAN_STD"
                }
            }
        },
        {
            "registry_name": "pi05_prepare_state_tokenizer_processor_step",
            "config": {}
        },
        {
            "registry_name": "tokenizer_processor",
            "config": {
                "max_length": 200,
                "task_key": "task",
                "padding_side": "right",
                "padding": "max_length",
                "truncation": True,
                "tokenizer_name": "google/paligemma-3b-pt-224"
            }
        },
        {
            "registry_name": "device_processor",
            "config": {
                "device": "cuda",
                "float_dtype": None
            }
        }
    ]
}

# Create postprocessor configuration (for unnormalizing actions)
postprocessor = {
    "name": "policy_postprocessor",
    "steps": [
        {
            "registry_name": "device_processor",
            "config": {
                "device": "cpu",
                "float_dtype": None
            }
        },
        {
            "registry_name": "unnormalizer_processor",
            "config": {
                "eps": 1e-08,
                "features": {
                    "action": features["action"]
                },
                "norm_map": {
                    "ACTION": "MEAN_STD"
                }
            }
        }
    ]
}

# Save configurations
output_dir = Path("pi05_pytorch")
with open(output_dir / "policy_preprocessor.json", "w") as f:
    json.dump(preprocessor, f, indent=2)

with open(output_dir / "policy_postprocessor.json", "w") as f:
    json.dump(postprocessor, f, indent=2)

print("✓ Created processor configurations:")
print(f"  - {output_dir / 'policy_preprocessor.json'}")
print(f"  - {output_dir / 'policy_postprocessor.json'}")
print("\nFeatures configured:")
for name, feature in features.items():
    print(f"  - {name}: {feature['type']} {feature['shape']}")