Python 3.10 support & bugfixing (#95)

* Python 3.10

* improve docs

* update req

* update README

* upgrade optuna

* bugfixing

* rework caching flows

* cleanup args

* update args

* cleanup benchmark args

* update tests

* add citation file

.github/workflows/test.yml

@@ -38,7 +38,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        python-version: ['3.7', '3.8', '3.9']
+        python-version: ['3.7', '3.8', '3.9', '3.10']
         os: [macos-latest, ubuntu-latest, windows-latest]
     steps:
       - uses: actions/checkout@v2

.github/workflows/test_tutorials.yml

@@ -14,7 +14,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        python-version: ["3.7", "3.8", "3.9"]
+        python-version: ["3.7", "3.8", "3.9", "3.10"]
         os: [ubuntu-latest]
     steps:
       - uses: actions/checkout@v2

CITATION.cff

@@ -0,0 +1,28 @@
+# This CITATION.cff file was generated with cffinit.
+# Visit https://bit.ly/cffinit to generate yours today!
+
+cff-version: 1.2.0
+title: >-
+  Synthcity: facilitating innovative use cases of synthetic
+  data in different data modalities
+message: >-
+  If you use this software, please cite it using the
+  metadata from this file.
+type: software
+authors:
+  - given-names: Zhaozhi
+    family-names: Qian
+  - given-names: Bogdan-Constantin
+    family-names: Cebere
+  - given-names: Mihaela
+    family-names: van der Schaar
+identifiers:
+  - type: doi
+    value: 10.48550/ARXIV.2301.07573
+repository-code: 'https://github.com/vanderschaarlab/synthcity'
+url: 'https://arxiv.org/abs/2301.07573'
+keywords:
+  - Machine Learning (cs.LG)
+  - Artificial Intelligence (cs.AI)
+  - 'FOS: Computer and information sciences'
+license: Apache-2.0

README.md

@@ -1,5 +1,5 @@
 <h2 align="center">
-  synthcity BETA
+  synthcity
 </h2>
 
 <h4 align="center">
@@ -18,6 +18,8 @@
 [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://github.com/vanderschaarlab/synthcity/blob/main/LICENSE)
 [![Python 3.7+](https://img.shields.io/badge/python-3.7+-blue.svg)](https://www.python.org/downloads/release/python-370/)
 
+![image](https://github.com/vanderschaarlab/synthcity/raw/main/docs/arch.png "Synthcity")
+
 </div>
 
 
@@ -33,8 +35,6 @@
 - :book: [Read the docs !](https://synthcity.readthedocs.io/)
 - :airplane: [Checkout the tutorials!](https://github.com/vanderschaarlab/synthcity#-tutorials)
 
-:rotating_light: __NOTE__: Python 3.10 is __NOT__ supported yet.
-
 ## :rocket: Installation
 
 The library can be installed from PyPI using
@@ -353,3 +353,18 @@ The tests can be executed using
 ```bash
 pytest -vsx
 ```
+## Citing
+
+If you use this code, please cite the associated paper:
+
+```
+@misc{https://doi.org/10.48550/arxiv.2301.07573,
+  doi = {10.48550/ARXIV.2301.07573},
+  url = {https://arxiv.org/abs/2301.07573},
+  author = {Qian, Zhaozhi and Cebere, Bogdan-Constantin and van der Schaar, Mihaela},
+  keywords = {Machine Learning (cs.LG), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
+  title = {Synthcity: facilitating innovative use cases of synthetic data in different data modalities},
+  year = {2023},
+  copyright = {Creative Commons Attribution 4.0 International}
+}
+```

setup.cfg

@@ -40,7 +40,8 @@ install_requires =
     lifelines>=0.27
     opacus>=1.3
     decaf-synthetic-data>=0.1.5
-    rdt>=1.2.1
+    rdt>=1.3
+    optuna>=3.1
     diffprivlib
     shap
     tqdm
@@ -53,7 +54,6 @@ install_requires =
     geomloss
     deepecho
     pgmpy
-    optuna
     redis
     pycox
     xgbse

src/synthcity/benchmark/__init__.py

@@ -103,6 +103,9 @@ def evaluate(
 
         for testcase, plugin, kwargs in tests:
             log.info(f"Testcase : {testcase}")
+            if not isinstance(kwargs, dict):
+                raise ValueError(f"'kwargs' must be a dict for {testcase}:{plugin}")
+
             scores = ScoreEvaluator()
 
             kwargs_hash = ""
@@ -116,6 +119,10 @@ def evaluate(
 
             for repeat in repeats_list:
                 enable_reproducible_results(repeat)
+
+                kwargs["workspace"] = workspace
+                kwargs["random_state"] = repeat
+
                 torch.cuda.empty_cache()
 
                 cache_file = (

src/synthcity/metrics/eval_detection.py

@@ -126,6 +126,9 @@ class SyntheticDetectionXGB(DetectionEvaluator):
         1: The datasets are totally distinguishable.
     """
 
+    def __init__(self, **kwargs: Any) -> None:
+        super().__init__(**kwargs)
+
     @staticmethod
     def name() -> str:
         return "detection_xgb"
@@ -158,6 +161,9 @@ class SyntheticDetectionMLP(DetectionEvaluator):
         1: The datasets are totally distinguishable.
     """
 
+    def __init__(self, **kwargs: Any) -> None:
+        super().__init__(**kwargs)
+
     @staticmethod
     def name() -> str:
         return "detection_mlp"
@@ -193,6 +199,9 @@ class SyntheticDetectionLinear(DetectionEvaluator):
         1: The datasets are totally distinguishable.
     """
 
+    def __init__(self, **kwargs: Any) -> None:
+        super().__init__(**kwargs)
+
     @staticmethod
     def name() -> str:
         return "detection_linear"
@@ -227,6 +236,9 @@ class SyntheticDetectionGMM(DetectionEvaluator):
         1: The datasets are totally distinguishable.
     """
 
+    def __init__(self, **kwargs: Any) -> None:
+        super().__init__(**kwargs)
+
     @staticmethod
     def name() -> str:
         return "detection_gmm"

src/synthcity/plugins/core/dataloader.py

@@ -245,11 +245,7 @@ def decode(
         decoded = self.dataframe().copy()
 
         for col in encoders:
-            decoded_local = decoded[[col]].copy()
-            decoded_local[f"{col}.value"] = decoded_local[col]
-            decoded[col] = encoders[col].reverse_transform(
-                decoded_local[[f"{col}.value"]].astype(float)
-            )
+            decoded[col] = encoders[col].reverse_transform(decoded[[col]].astype(float))
 
         return self.from_info(decoded, self.info())
 

src/synthcity/plugins/core/models/tabular_encoder.py

@@ -4,9 +4,7 @@
 """
 
 # stdlib
-import platform
 from collections import namedtuple
-from pathlib import Path
 from typing import Any, List, Optional, Sequence, Tuple
 
 # third party
@@ -19,7 +17,7 @@
 
 # synthcity absolute
 import synthcity.logger as log
-from synthcity.utils.serialization import dataframe_hash, load_from_file, save_to_file
+from synthcity.utils.serialization import dataframe_hash
 
 ColumnTransformInfo = namedtuple(
     "ColumnTransformInfo",
@@ -67,7 +65,7 @@ def _fit_continuous(self, data: pd.Series) -> ColumnTransformInfo:
             max_clusters=min(self.max_clusters, len(data)),
             enforce_min_max_values=True,
         )
-        gm.fit(data.to_frame(), [column_name])
+        gm.fit(data.to_frame(), column_name)
         num_components = sum(gm.valid_component_indicator)
 
         output_columns = [f"{column_name}.normalized"] + [
@@ -111,7 +109,7 @@ def _transform_continuous(
     ) -> pd.Series:
         column_name = data.name
         gm = column_transform_info.transform
-        transformed = gm.transform(data.to_frame(), [column_name])
+        transformed = gm.transform(data.to_frame()[[column_name]])
 
         return transformed[f"{column_name}.component"].to_numpy().astype(int)
 
@@ -147,7 +145,6 @@ def __init__(
         max_clusters: int = 10,
         categorical_limit: int = 10,
         whitelist: list = [],
-        workspace: Path = Path("workspace"),
     ) -> None:
         """Create a data transformer.
 
@@ -158,8 +155,6 @@ def __init__(
         self.max_clusters = max_clusters
         self.categorical_limit = categorical_limit
         self.whitelist = whitelist
-        self.workspace = workspace
-        self.workspace.mkdir(parents=True, exist_ok=True)
 
     @validate_arguments(config=dict(arbitrary_types_allowed=True))
     def _fit_continuous(self, data: pd.Series) -> ColumnTransformInfo:
@@ -179,7 +174,7 @@ def _fit_continuous(self, data: pd.Series) -> ColumnTransformInfo:
             max_clusters=min(len(data), self.max_clusters),
             enforce_min_max_values=True,
         )
-        gm.fit(data.to_frame(), [column_name])
+        gm.fit(data.to_frame(), column_name)
         num_components = sum(gm.valid_component_indicator)
 
         output_columns = [f"{column_name}.normalized"] + [
@@ -243,27 +238,16 @@ def fit(
         self._column_raw_dtypes = raw_data.infer_objects().dtypes
         self._column_transform_info_list = []
 
-        self.workspace.mkdir(parents=True, exist_ok=True)
-
         for column_name in raw_data.columns:
             if column_name in self.whitelist:
                 continue
             column_hash = dataframe_hash(raw_data[[column_name]])
-            bkp_file = (
-                self.workspace
-                / f"encoder_cache_{column_hash}_{column_name[:50]}_{self.max_clusters}_{self.categorical_limit}_{platform.python_version()}.bkp"
-            )
-
             log.info(f"Encoding {column_name} {column_hash}")
 
-            if bkp_file.exists():
-                column_transform_info = load_from_file(bkp_file)
+            if column_name in discrete_columns:
+                column_transform_info = self._fit_discrete(raw_data[column_name])
             else:
-                if column_name in discrete_columns:
-                    column_transform_info = self._fit_discrete(raw_data[column_name])
-                else:
-                    column_transform_info = self._fit_continuous(raw_data[column_name])
-                save_to_file(bkp_file, column_transform_info)
+                column_transform_info = self._fit_continuous(raw_data[column_name])
 
             self.output_dimensions += column_transform_info.output_dimensions
             self._column_transform_info_list.append(column_transform_info)
@@ -275,7 +259,7 @@ def _transform_continuous(
     ) -> pd.DataFrame:
         column_name = data.name
         gm = column_transform_info.transform
-        transformed = gm.transform(data.to_frame(), [column_name])
+        transformed = gm.transform(data.to_frame()[[column_name]])
 
         #  Converts the transformed data to the appropriate output format.
         #  The first column (ending in '.normalized') stays the same,
@@ -341,7 +325,7 @@ def _inverse_transform_continuous(
             column_data.values[:, :2], columns=list(gm.get_output_sdtypes())
         )
         data.iloc[:, 1] = np.argmax(column_data.values[:, 1:], axis=1)
-        return gm.reverse_transform(data, [column_transform_info.column_name])
+        return gm.reverse_transform(data)
 
     @validate_arguments(config=dict(arbitrary_types_allowed=True))
     def _inverse_transform_discrete(

src/synthcity/plugins/core/plugin.py

@@ -35,6 +35,7 @@
 from synthcity.plugins.core.schema import Schema
 from synthcity.plugins.core.serializable import Serializable
 from synthcity.utils.constants import DEVICE
+from synthcity.utils.reproducibility import enable_reproducible_results
 from synthcity.utils.serialization import load_from_file, save_to_file
 
 
@@ -86,6 +87,9 @@ def __init__(
         sampling_strategy: str = "marginal",  # uniform, marginal
     ) -> None:
         super().__init__()
+
+        enable_reproducible_results(random_state)
+
         self._schema: Optional[Schema] = None
         self._training_schema: Optional[Schema] = None
         self._data_encoders: Optional[Dict] = None

src/synthcity/plugins/generic/plugin_bayesian_network.py

@@ -4,6 +4,7 @@
 """
 
 # stdlib
+from pathlib import Path
 from typing import Any, List
 
 # third party
@@ -38,9 +39,16 @@ class BayesianNetworkPlugin(Plugin):
             The maximum number of parents for each node.
         encoder_max_clusters: int = 10
             Data encoding clusters.
-        encoder_noise_scale: float
+        encoder_noise_scale: float.
             Small noise to add to the final data, to prevent data leakage.
-
+        workspace: Path.
+            Optional Path for caching intermediary results.
+        compress_dataset: bool. Default = False.
+            Drop redundant features before training the generator.
+        random_state: int.
+            Random seed.
+        sampling_patience: int.
+            Max inference iterations to wait for the generated data to match the training schema.
 
     Example:
         >>> from sklearn.datasets import load_iris
@@ -64,9 +72,20 @@ def __init__(
         struct_max_indegree: int = 4,
         encoder_max_clusters: int = 10,
         encoder_noise_scale: float = 0.1,
+        # core plugin
+        workspace: Path = Path("workspace"),
+        compress_dataset: bool = False,
+        random_state: int = 0,
+        sampling_patience: int = 500,
         **kwargs: Any,
     ) -> None:
-        super().__init__(**kwargs)
+        super().__init__(
+            random_state=random_state,
+            sampling_patience=sampling_patience,
+            workspace=workspace,
+            compress_dataset=compress_dataset,
+            **kwargs,
+        )
 
         self.struct_learning_n_iter = struct_learning_n_iter
         self.struct_learning_search_method = struct_learning_search_method