sdgym 0.9.0.dev0__tar.gz → 0.9.1.dev0__tar.gz

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sdgym
-Version: 0.9.0.dev0
+Version: 0.9.1.dev0
 Summary: Benchmark tabular synthetic data generators using a variety of datasets
 Author-email: "DataCebo, Inc." <info@sdv.dev>
 License: BSL-1.1
@@ -27,6 +27,7 @@ License-File: LICENSE
 Requires-Dist: appdirs>=1.3
 Requires-Dist: boto3<2,>=1.28
 Requires-Dist: botocore<2,>=1.31
+Requires-Dist: cloudpickle>=2.1.0
 Requires-Dist: compress-pickle>=1.2.0
 Requires-Dist: humanfriendly>=8.2
 Requires-Dist: numpy<2.0.0,>=1.21.0; python_version < "3.10"

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/pyproject.toml

@@ -24,6 +24,7 @@ dependencies = [
     'appdirs>=1.3',
     'boto3>=1.28,<2',
     'botocore>=1.31,<2',
+    'cloudpickle>=2.1.0',
     'compress-pickle>=1.2.0',
     'humanfriendly>=8.2',
     "numpy>=1.21.0,<2.0.0;python_version<'3.10'",
@@ -133,7 +134,7 @@ namespaces = false
 version = {attr = 'sdgym.__version__'}
 
 [tool.bumpversion]
-current_version = "0.9.0.dev0"
+current_version = "0.9.1.dev0"
 parse = '(?P<major>\d+)\.(?P<minor>\d+)\.(?P<patch>\d+)(\.(?P<release>[a-z]+)(?P<candidate>\d+))?'
 serialize = [
     '{major}.{minor}.{patch}.{release}{candidate}',

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym/__init__.py

@@ -8,7 +8,7 @@ __author__ = 'DataCebo, Inc.'
 __copyright__ = 'Copyright (c) 2022 DataCebo, Inc.'
 __email__ = 'info@sdv.dev'
 __license__ = 'BSL-1.1'
-__version__ = '0.9.0.dev0'
+__version__ = '0.9.1.dev0'
 
 import logging
 

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym/benchmark.py

@@ -7,10 +7,12 @@ import os
 import pickle
 import tracemalloc
 import warnings
+from contextlib import contextmanager
 from datetime import datetime
 from pathlib import Path
 
 import boto3
+import cloudpickle
 import compress_pickle
 import numpy as np
 import pandas as pd
@@ -146,11 +148,18 @@ def _generate_job_args_list(
 
 def _synthesize(synthesizer_dict, real_data, metadata):
     synthesizer = synthesizer_dict['synthesizer']
-    assert issubclass(synthesizer, BaselineSynthesizer), '`synthesizer` must be a synthesizer class'
+    if isinstance(synthesizer, type):
+        assert issubclass(
+            synthesizer, BaselineSynthesizer
+        ), '`synthesizer` must be a synthesizer class'
+        synthesizer = synthesizer()
+    else:
+        assert issubclass(
+            type(synthesizer), BaselineSynthesizer
+        ), '`synthesizer` must be an instance of a synthesizer class.'
 
-    synthesizer_object = synthesizer()
-    get_synthesizer = synthesizer_object.get_trained_synthesizer
-    sample_from_synthesizer = synthesizer_object.sample_from_synthesizer
+    get_synthesizer = synthesizer.get_trained_synthesizer
+    sample_from_synthesizer = synthesizer.sample_from_synthesizer
     data = real_data.copy()
     num_samples = len(data)
 
@@ -190,7 +199,8 @@ def _compute_scores(
                 'metric': metric_name,
                 'error': 'Metric Timeout',
             })
-            output['scores'] = scores  # re-inject list to multiprocessing output
+            # re-inject list to multiprocessing output
+            output['scores'] = scores
 
             error = None
             score = None
@@ -213,7 +223,8 @@ def _compute_scores(
                 'error': error,
                 'metric_time': (datetime.utcnow() - start).total_seconds(),
             })
-            output['scores'] = scores  # re-inject list to multiprocessing output
+            # re-inject list to multiprocessing output
+            output['scores'] = scores
 
     if compute_diagnostic_score:
         start = datetime.utcnow()
@@ -264,7 +275,8 @@ def _score(
         )
 
         output['dataset_size'] = get_size_of(data) / N_BYTES_IN_MB
-        output['error'] = 'Synthesizer Timeout'  # To be deleted if there is no error
+        # To be deleted if there is no error
+        output['error'] = 'Synthesizer Timeout'
         synthetic_data, train_time, sample_time, synthesizer_size, peak_memory = _synthesize(
             synthesizer, data.copy(), metadata
         )
@@ -283,7 +295,8 @@ def _score(
             used_memory(),
         )
 
-        del output['error']  # No error so far. _compute_scores tracks its own errors by metric
+        # No error so far. _compute_scores tracks its own errors by metric
+        del output['error']
         _compute_scores(
             metrics,
             data,
@@ -314,6 +327,26 @@ def _score(
     return output
 
 
+@contextmanager
+def multiprocessing_context():
+    """Override multiprocessing ForkingPickler to use cloudpickle."""
+    original_dump = multiprocessing.reduction.ForkingPickler.dumps
+    original_load = multiprocessing.reduction.ForkingPickler.loads
+    original_method = multiprocessing.get_start_method()
+
+    multiprocessing.set_start_method('spawn', force=True)
+    multiprocessing.reduction.ForkingPickler.dumps = cloudpickle.dumps
+    multiprocessing.reduction.ForkingPickler.loads = cloudpickle.loads
+
+    try:
+        yield
+    finally:
+        # Restore original methods
+        multiprocessing.set_start_method(original_method, force=True)
+        multiprocessing.reduction.ForkingPickler.dumps = original_dump
+        multiprocessing.reduction.ForkingPickler.loads = original_load
+
+
 def _score_with_timeout(
     timeout,
     synthesizer,
@@ -325,32 +358,33 @@ def _score_with_timeout(
     modality=None,
     dataset_name=None,
 ):
-    with multiprocessing.Manager() as manager:
-        output = manager.dict()
-        process = multiprocessing.Process(
-            target=_score,
-            args=(
-                synthesizer,
-                data,
-                metadata,
-                metrics,
-                output,
-                compute_quality_score,
-                compute_diagnostic_score,
-                modality,
-                dataset_name,
-            ),
-        )
+    with multiprocessing_context():
+        with multiprocessing.Manager() as manager:
+            output = manager.dict()
+            process = multiprocessing.Process(
+                target=_score,
+                args=(
+                    synthesizer,
+                    data,
+                    metadata,
+                    metrics,
+                    output,
+                    compute_quality_score,
+                    compute_diagnostic_score,
+                    modality,
+                    dataset_name,
+                ),
+            )
 
-        process.start()
-        process.join(timeout)
-        process.terminate()
+            process.start()
+            process.join(timeout)
+            process.terminate()
 
-        output = dict(output)
-        if output.get('timeout'):
-            LOGGER.error('Timeout running %s on dataset %s;', synthesizer['name'], dataset_name)
+            output = dict(output)
+            if output.get('timeout'):
+                LOGGER.error('Timeout running %s on dataset %s;', synthesizer['name'], dataset_name)
 
-        return output
+            return output
 
 
 def _format_output(
@@ -756,6 +790,11 @@ def benchmark_single_table(
              'package_name': 'dask' or 'multiprocessing',
              'num_workers': 4
             }
+        run_on_ec2 (bool):
+            The flag is used to run the benchmark on an EC2 instance that will be created
+            by a scriptusing the authentication of the current user. The EC2 instance
+            uses the LATEST released version of sdgym. Local changes or changes NOT
+            in the released version will NOT be used in the ec2 instance.
 
     Returns:
         pandas.DataFrame:

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym/synthesizers/generate.py

@@ -124,7 +124,7 @@ def create_single_table_synthesizer(
                 obj:
                     The trained synthesizer.
             """
-            return get_trained_synthesizer_fn(data, metadata)
+            return self.synthesizer_fn['get_trained_synthesizer_fn'](data, metadata)
 
         def sample_from_synthesizer(self, synthesizer, num_samples):
             """Sample the desired number of samples from the given synthesizer.
@@ -139,11 +139,22 @@ def create_single_table_synthesizer(
                 pandas.DataFrame:
                     The synthetic data.
             """
-            return sample_from_synthesizer_fn(synthesizer, num_samples)
-
-    NewSynthesizer.__name__ = f'Custom:{display_name}'
-
-    return NewSynthesizer
+            return self.synthesizer_fn['sample_from_synthesizer_fn'](synthesizer, num_samples)
+
+    CustomSynthesizer = type(
+        f'Custom:{display_name}',
+        (NewSynthesizer,),
+        {
+            'synthesizer_fn': {
+                'get_trained_synthesizer_fn': get_trained_synthesizer_fn,
+                'sample_from_synthesizer_fn': sample_from_synthesizer_fn,
+            },
+        },
+    )
+    CustomSynthesizer.__name__ = f'Custom:{display_name}'
+    CustomSynthesizer.__module__ = 'sdgym.synthesizers.generate'
+    globals()[f'Custom:{display_name}'] = CustomSynthesizer
+    return CustomSynthesizer
 
 
 def create_multi_table_synthesizer(

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym/utils.py

@@ -60,8 +60,12 @@ def get_synthesizers(synthesizers):
             else:
                 raise SDGymError(f'Unknown synthesizer {synthesizer}') from None
 
+        if isinstance(synthesizer, type) or hasattr(synthesizer, '__name__'):
+            synthesizer_name = getattr(synthesizer, '__name__', 'undefined')
+        else:
+            synthesizer_name = getattr(type(synthesizer), '__name__', 'undefined')
         synthesizers_dicts.append({
-            'name': getattr(synthesizer, '__name__', 'undefined'),
+            'name': synthesizer_name,
             'synthesizer': synthesizer,
         })
 

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sdgym
-Version: 0.9.0.dev0
+Version: 0.9.1.dev0
 Summary: Benchmark tabular synthetic data generators using a variety of datasets
 Author-email: "DataCebo, Inc." <info@sdv.dev>
 License: BSL-1.1
@@ -27,6 +27,7 @@ License-File: LICENSE
 Requires-Dist: appdirs>=1.3
 Requires-Dist: boto3<2,>=1.28
 Requires-Dist: botocore<2,>=1.31
+Requires-Dist: cloudpickle>=2.1.0
 Requires-Dist: compress-pickle>=1.2.0
 Requires-Dist: humanfriendly>=8.2
 Requires-Dist: numpy<2.0.0,>=1.21.0; python_version < "3.10"

{sdgym-0.9.0.dev0 → sdgym-0.9.1.dev0}/sdgym.egg-info/requires.txt

@@ -1,6 +1,7 @@
 appdirs>=1.3
 boto3<2,>=1.28
 botocore<2,>=1.31
+cloudpickle>=2.1.0
 compress-pickle>=1.2.0
 humanfriendly>=8.2
 psutil>=5.7