gitea_admin/sientia-dataops-model-manager
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SIENTIAPDE-1251: Implement ML model training activity and repository

Browse Source 
This commit introduces the 'Training' activity and 'TrainingRepository' for handling ML model training operations within the Model Manager system.

- Added model_manager/activities/training.py for the Training activity, which extends BaseActivity and integrates with Temporal workflows.
- Added model_manager/utils/repository/training_repository.py for the TrainingRepository, which encapsulates the core training logic.
- Updated model_manager/activities/activities.py to include the Training activity in the main activities orchestrator.
- Updated README.md to document the new 'Training' component.
- Added unit tests for the new activity and repository.
This commit is contained in:
Bruno Domingues
2025-10-09 15:07:27 -03:00
parent c970754e1b
commit bee6036205
7 changed files with 1046 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

327
tests/utils/repository/test_training_repository.py Normal file
View File

@@ -0,0 +1,327 @@
"""Unit tests for TrainingRepository."""
from io import BytesIO
from unittest.mock import MagicMock, patch
import numpy as np
import pandas as pd
from pytest import fixture, raises
from model_manager.utils.models.train_model_params import TrainModelParams
from model_manager.utils.models.train_model_result import TrainModelResult
from model_manager.utils.repository.training_repository import TrainingRepository
@fixture
def logger():
"""Create a mock logger."""
return MagicMock()
@fixture
def training_repository(logger):
"""Create a TrainingRepository instance."""
return TrainingRepository(logger)
@fixture
def train_params():
"""Create sample training parameters."""
return TrainModelParams(
variable_columns=['feature1', 'feature2'],
lag_train=1,
lag_val=1,
target_variable='target',
rem_static_win=False,
low_lim={'feature1': 0.0, 'feature2': 0.0},
upp_lim={'feature1': 100.0, 'feature2': 100.0},
window=10,
use_scaler=True,
include_ar=False,
bucket_name='test-bucket',
file_name='test.csv',
line_separator='\n',
decimal_separator='.',
train_size=80,
shuffle=True,
experiment_run_id=123,
experiment_name='test_experiment',
experiment_description='Test experiment',
removed_intervals=[],
)
@fixture
def sample_csv_data():
"""Create sample CSV data."""
csv_content = """feature1,feature2,target
1.0,2.0,10.0
2.0,3.0,15.0
3.0,4.0,20.0
4.0,5.0,25.0
5.0,6.0,30.0
6.0,7.0,35.0
7.0,8.0,40.0
8.0,9.0,45.0
9.0,10.0,50.0
10.0,11.0,55.0
"""
return BytesIO(csv_content.encode())
@patch('model_manager.utils.repository.training_repository.load_data')
@patch('model_manager.utils.repository.training_repository.DataPreprocessor')
@patch('model_manager.utils.repository.training_repository.split_train_test')
@patch('model_manager.utils.repository.training_repository.LinearRegressionModel')
def test_train_success(
mock_linear_model,
mock_split,
mock_preprocessor_class,
mock_load_data,
training_repository,
train_params,
sample_csv_data,
):
"""Test successful model training."""
# Setup mocks
mock_data = pd.DataFrame(
{'feature1': [1, 2, 3, 4, 5], 'feature2': [2, 3, 4, 5, 6], 'target': [10, 15, 20, 25, 30]}
)
mock_load_data.return_value = mock_data
mock_preprocessor = MagicMock()
mock_preprocessor_class.return_value = mock_preprocessor
mock_preprocessor.transform.return_value = mock_data
x_train = pd.DataFrame({'feature1': [1, 2, 3], 'feature2': [2, 3, 4]})
x_test = pd.DataFrame({'feature1': [4, 5], 'feature2': [5, 6]})
y_train = pd.Series([10, 15, 20], name='target')
y_test = pd.Series([25, 30], name='target')
mock_split.return_value = (x_train, x_test, y_train, y_test)
mock_model = MagicMock()
mock_linear_model.return_value = mock_model
mock_scaler = MagicMock()
mock_preprocessor.get_scaler.return_value = mock_scaler
# Execute
result = training_repository.train(sample_csv_data, train_params)
# Assertions
assert isinstance(result, TrainModelResult)
assert result.params == train_params
assert result.process_data == mock_preprocessor
assert result.regr == mock_model
mock_load_data.assert_called_once()
mock_preprocessor.fit.assert_called_once()
mock_model.fit.assert_called_once()
@patch('model_manager.utils.repository.training_repository.load_data')
def test_train_empty_data_after_transform(
mock_load_data, training_repository, train_params, sample_csv_data
):
"""Test training with empty data after transformation."""
mock_data = pd.DataFrame({'feature1': [], 'feature2': [], 'target': []})
mock_load_data.return_value = mock_data
with patch.object(training_repository, 'init_data_preprocessor') as mock_init:
mock_preprocessor = MagicMock()
mock_init.return_value = mock_preprocessor
mock_preprocessor.transform.return_value = pd.DataFrame()
with raises(ValueError, match='Data view is empty after transformation'):
training_repository.train(sample_csv_data, train_params)
def test_init_scaler_dict_with_minmax_scaler(training_repository, train_params):
"""Test scaler dict initialization with MinMaxScaler."""
mock_preprocessor = MagicMock()
mock_scaler = MagicMock()
mock_scaler.x_min = [0.0, 1.0]
mock_scaler.x_max = [10.0, 11.0]
mock_scaler.y_min = 5.0
mock_scaler.y_max = 50.0
mock_preprocessor.get_scaler.return_value = mock_scaler
# Patch isinstance to return True for MinMaxScaler
with patch(
'model_manager.utils.repository.training_repository.isinstance',
side_effect=lambda obj, cls: cls.__name__ == 'MinMaxScaler',
):
result = training_repository.init_scaler_dict(mock_preprocessor, train_params)
assert result is not None
assert 'feature1' in result
assert 'feature2' in result
assert 'target' in result
assert result['feature1'] == {'min': 0.0, 'max': 10.0}
assert result['feature2'] == {'min': 1.0, 'max': 11.0}
assert result['target'] == {'min': 5.0, 'max': 50.0}
def test_init_scaler_dict_with_z_scaler(training_repository, train_params):
"""Test scaler dict initialization with Z_Scaler."""
mock_preprocessor = MagicMock()
mock_scaler = MagicMock()
mock_scaler.create_dict.return_value = {'mean': 5.0, 'std': 2.0}
mock_preprocessor.get_scaler.return_value = mock_scaler
# Patch isinstance to return True for Z_Scaler
with patch(
'model_manager.utils.repository.training_repository.isinstance',
side_effect=lambda obj, cls: cls.__name__ == 'Z_Scaler',
):
result = training_repository.init_scaler_dict(mock_preprocessor, train_params)
assert result == {'mean': 5.0, 'std': 2.0}
mock_scaler.create_dict.assert_called_once()
def test_init_scaler_dict_without_scaler(training_repository):
"""Test scaler dict initialization when use_scaler is False."""
train_params_no_scaler = TrainModelParams(
variable_columns=['feature1'],
lag_train=1,
lag_val=1,
target_variable='target',
rem_static_win=False,
low_lim={'feature1': 0.0},
upp_lim={'feature1': 100.0},
window=10,
use_scaler=False,
include_ar=False,
bucket_name='test',
file_name='test.csv',
line_separator='\n',
decimal_separator='.',
train_size=80,
shuffle=True,
experiment_run_id=123,
experiment_name='test',
experiment_description='test',
removed_intervals=[],
)
mock_preprocessor = MagicMock()
result = training_repository.init_scaler_dict(mock_preprocessor, train_params_no_scaler)
assert result == {}
@patch('model_manager.utils.repository.training_repository.mse')
@patch('model_manager.utils.repository.training_repository.mae')
@patch('model_manager.utils.repository.training_repository.r2')
def test_after_train_calculation_with_scaler(
mock_r2, mock_mae, mock_mse, training_repository, train_params
):
"""Test post-training calculations with scaler."""
# Setup mock train result
mock_train_result = MagicMock(spec=TrainModelResult)
mock_train_result.params = train_params
mock_train_result.x_train = pd.DataFrame({'feature1': [1, 2, 3], 'feature2': [2, 3, 4]})
mock_train_result.x_test = pd.DataFrame({'feature1': [4, 5], 'feature2': [5, 6]})
mock_train_result.y_train = pd.Series([10, 15, 20], name='target')
mock_train_result.y_test = pd.Series([25, 30], name='target')
mock_regr = MagicMock()
mock_regr.predict.return_value = np.array([24.5, 29.5])
mock_train_result.regr = mock_regr
mock_scaler = MagicMock()
mock_scaler.denormalize_single_input.side_effect = lambda x, col: x
mock_scaler.denormalize_predictions.side_effect = lambda x, col: x
mock_process_data = MagicMock()
mock_process_data.get_scaler.return_value = mock_scaler
mock_train_result.process_data = mock_process_data
# Setup metric mocks
mock_mse.return_value = 0.5
mock_mae.return_value = 0.3
mock_r2.return_value = 0.95
# Execute
result = training_repository.after_train_calculation(train_params, mock_train_result)
# Assertions
assert result == mock_train_result
assert result.mse_val == 0.5
assert result.mae_val == 0.3
assert result.r2_val == 0.95
assert result.y_pred is not None
mock_regr.predict.assert_called_once()
mock_mse.assert_called_once()
mock_mae.assert_called_once()
mock_r2.assert_called_once()
@patch('model_manager.utils.repository.training_repository.mse')
@patch('model_manager.utils.repository.training_repository.mae')
@patch('model_manager.utils.repository.training_repository.r2')
def test_after_train_calculation_without_scaler(mock_r2, mock_mae, mock_mse, training_repository):
"""Test post-training calculations without scaler."""
train_params_no_scaler = TrainModelParams(
variable_columns=['feature1'],
lag_train=1,
lag_val=1,
target_variable='target',
rem_static_win=False,
low_lim={'feature1': 0.0},
upp_lim={'feature1': 100.0},
window=10,
use_scaler=False,
include_ar=False,
bucket_name='test',
file_name='test.csv',
line_separator='\n',
decimal_separator='.',
train_size=80,
shuffle=True,
experiment_run_id=123,
experiment_name='test',
experiment_description='test',
removed_intervals=[],
)
mock_train_result = MagicMock(spec=TrainModelResult)
mock_train_result.params = train_params_no_scaler
mock_train_result.x_train = pd.DataFrame({'feature1': [1, 2, 3]})
mock_train_result.x_test = pd.DataFrame({'feature1': [4, 5]})
mock_train_result.y_train = pd.Series([10, 15, 20], name='target')
mock_train_result.y_test = pd.Series([25, 30], name='target')
mock_regr = MagicMock()
mock_regr.predict.return_value = np.array([24.5, 29.5])
mock_train_result.regr = mock_regr
# Setup metric mocks
mock_mse.return_value = 0.5
mock_mae.return_value = 0.3
mock_r2.return_value = 0.95
# Execute
result = training_repository.after_train_calculation(train_params_no_scaler, mock_train_result)
# Assertions
assert result.mse_val == 0.5
assert result.mae_val == 0.3
assert result.r2_val == 0.95
@patch('model_manager.utils.repository.training_repository.DataPreprocessor')
def test_init_data_preprocessor(mock_preprocessor_class, training_repository, train_params):
"""Test DataPreprocessor initialization."""
mock_preprocessor = MagicMock()
mock_preprocessor_class.return_value = mock_preprocessor
result = training_repository.init_data_preprocessor(train_params)
assert result == mock_preprocessor
mock_preprocessor_class.assert_called_once()
call_kwargs = mock_preprocessor_class.call_args[1]
assert call_kwargs['target_variable'] == 'target'
assert call_kwargs['input_columns'] == ['feature1', 'feature2']
assert call_kwargs['low_lim'] == {'feature1': 0.0, 'feature2': 0.0}
assert call_kwargs['upp_lim'] == {'feature1': 100.0, 'feature2': 100.0}