perpetual 2.0.0

A self-generalizing gradient boosting machine that doesn't need hyperparameter optimization
Documentation 
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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209

import os
import subprocess

import pandas as pd
import seaborn as sns
from perpetual.utils import convert_input_frame, transform_input_frame
from sklearn.datasets import fetch_california_housing, fetch_covtype, fetch_openml
from sklearn.model_selection import train_test_split

if __name__ == "__main__":
    df = sns.load_dataset("titanic")
    df.to_csv("resources/titanic.csv", index=False)

    X = df.select_dtypes("number").drop(columns=["survived"]).astype(float)
    y = df["survived"].astype(float)

    pd.Series(X.fillna(0).to_numpy().ravel(order="F")).to_csv(
        "resources/contiguous_no_missing.csv",
        index=False,
        header=False,
    )

    pd.Series(X.to_numpy().ravel(order="F")).to_csv(
        "resources/contiguous_with_missing.csv",
        index=False,
        header=False,
    )

    y.to_csv(
        "resources/performance.csv",
        index=False,
        header=False,
    )

    X.fare.to_csv(
        "resources/performance-fare.csv",
        index=False,
        header=False,
    )

    dfb = df.sample(
        100_000,
        random_state=0,
        replace=True,
    ).reset_index(drop=True)

    Xb = dfb.select_dtypes("number").drop(columns=["survived"]).astype(float)
    yb = dfb["survived"].astype(float)

    pd.Series(Xb.fillna(0).to_numpy().ravel(order="F")).to_csv(
        "resources/contiguous_no_missing_100k_samp_seed0.csv",
        index=False,
        header=False,
    )

    yb.to_csv(
        "resources/performance_100k_samp_seed0.csv",
        index=False,
        header=False,
    )

    data = fetch_california_housing(as_frame=True)
    data_train, data_test = train_test_split(data.frame, test_size=0.2, random_state=42)
    data_train.to_csv("resources/cal_housing_train.csv", index=False)
    data_test.to_csv("resources/cal_housing_test.csv", index=False)

    data = fetch_covtype(as_frame=True)
    data_train, data_test = train_test_split(data.frame, test_size=0.2, random_state=42)
    data_train.to_csv("resources/cover_types_train.csv", index=False)
    data_test.to_csv("resources/cover_types_test.csv", index=False)

    X = df.drop(columns=["survived"])
    y = df["survived"]

    X["sex"] = pd.get_dummies(X["sex"], drop_first=True, dtype=float).to_numpy()
    X["adult_male"] = pd.get_dummies(
        X["adult_male"], drop_first=True, dtype=float
    ).to_numpy()
    X.drop(columns=["alive"], inplace=True)
    X["alone"] = pd.get_dummies(X["alone"], drop_first=True, dtype=float).to_numpy()
    cols = [
        "pclass",
        "sibsp",
        "parch",
        "embarked",
        "class",
        "who",
        "deck",
        "embark_town",
    ]
    X[cols] = X[cols].astype("category")

    data_train, data_test, y_train, y_test = train_test_split(
        X, y, test_size=0.2, random_state=42
    )

    (
        features_,
        titanic_train_flat,
        rows,
        cols,
        categorical_features_,
        cat_mapping,
    ) = convert_input_frame(data_train, "auto", 1000)
    features_, titanic_test_flat, rows, cols = transform_input_frame(
        data_test, cat_mapping
    )

    data_test.to_csv("resources/titanic_test_df.csv", index=False)

    pd.Series(titanic_train_flat).to_csv(
        "resources/titanic_train_flat.csv", index=False, header=False
    )
    pd.Series(titanic_test_flat).to_csv(
        "resources/titanic_test_flat.csv", index=False, header=False
    )
    pd.Series(y_train).to_csv(
        "resources/titanic_train_y.csv", index=False, header=False
    )
    pd.Series(y_test).to_csv("resources/titanic_test_y.csv", index=False, header=False)

    # https://www.openml.org/search?type=data&id=546&sort=runs&status=active
    df = fetch_openml(data_id=546)
    X = df.data
    y = df.target
    (
        features_,
        sensory_flat,
        rows,
        cols,
        categorical_features_,
        cat_mapping,
    ) = convert_input_frame(X, "auto", 1000)
    pd.Series(sensory_flat).to_csv(
        "resources/sensory_flat.csv", index=False, header=False
    )
    pd.Series(y).to_csv("resources/sensory_y.csv", index=False, header=False)

    # https://www.openml.org/search?type=data&id=43493&sort=runs&status=active
    df = fetch_openml(data_id=43493)
    cols_to_drop = ["title", "authors", "votes"]
    data = df.data.drop(columns=cols_to_drop)

    data["category"] = data["category"].astype("category")
    data["published"] = data["published"].astype("category")

    data.to_csv("resources/goodreads.csv", index=False, header=True)

    # ---------------------------------------------------------
    # Generate v2.0.0 Model Artifact for Backward Compatibility
    # ---------------------------------------------------------
    print("Generating v2.0.0 model artifact...")

    gen_script = """
import pandas as pd
import seaborn as sns
from sklearn.model_selection import train_test_split
from perpetual import PerpetualBooster

# Replicate Titanic data loading
df = sns.load_dataset("titanic")
X = df.drop(columns=["survived"])
y = df["survived"]
X["sex"] = pd.get_dummies(X["sex"], drop_first=True, dtype=float).to_numpy()
X["adult_male"] = pd.get_dummies(X["adult_male"], drop_first=True, dtype=float).to_numpy()
X.drop(columns=["alive"], inplace=True)
X["alone"] = pd.get_dummies(X["alone"], drop_first=True, dtype=float).to_numpy()
cols = ['pclass', 'sibsp', 'parch', 'embarked', 'class', 'who', 'deck', 'embark_town']
X[cols] = X[cols].astype('category')

data_train, data_test, y_train, _ = train_test_split(X, y, test_size=0.2, random_state=42)

model = PerpetualBooster(objective="LogLoss")
model.fit(data_train, y_train)
model.save_booster("resources/model_v2.0.0.json")

# Save predictions for verification
preds = model.predict(data_test)
pd.Series(preds).to_csv("resources/model_v2.0.0_preds.csv", index=False, header=False)
probs = model.predict_proba(data_test)
pd.DataFrame(probs).to_csv("resources/model_v2.0.0_probs.csv", index=False, header=False)
print("Successfully generated resources/model_v2.0.0.json and predictions")
"""
    with open("temp_gen_model.py", "w") as f:
        f.write(gen_script)

    try:
        subprocess.check_call(
            [
                "uv",
                "run",
                "--with",
                "./package-python",  # "./package-python" or "perpetual==2.0.0"
                "--with",
                "pandas",
                "--with",
                "seaborn",
                "--with",
                "scikit-learn",
                "python",
                "temp_gen_model.py",
            ]
        )
    except subprocess.CalledProcessError as e:
        print(f"Failed to generate model: {e}")
        raise e
    finally:
        if os.path.exists("temp_gen_model.py"):
            os.remove("temp_gen_model.py")