def run_experiment(project_id: int, engine: str, model: object,
target_col: str, train_data: pd.DataFrame,
test_data: pd.DataFrame, metrics: dict) -> int:
_ENGINES = ['sklearn']
if engine not in _ENGINES:
raise Exception(
f'Engine not registered, must be one of the following: {",".join(_ENGINES)}'
)
database = Database()
query = f"INSERT INTO experiment (project_id, engine) VALUES ('{project_id}', '{engine}')"
experiment_id = database.write(query=query)
registry = Registry()
registry.put_model(path=f"{project_id}-{experiment_id}",
key='pre-model',
model=model)
registry.put_metrics(path=f"{project_id}-{experiment_id}",
key='metrics',
metrics=metrics)
registry.put_dataset(path=f"{project_id}-{experiment_id}",
key='test',
dataset=test_data)
registry.put_dataset(path=f"{project_id}-{experiment_id}",
key='train',
dataset=train_data)
pipeline = Pipeline()
conf = {
'experiment_id': experiment_id,
'project_id': project_id,
'target_col': target_col
}
triggered = pipeline.trigger_dag(dag_id=f'{engine}-pipeline',
data=dict(conf=conf))
if triggered:
query = f"UPDATE experiment SET status = 'submitted' WHERE id = {experiment_id}"
database.write(query=query)
else:
query = f"UPDATE experiment SET status = 'submission-failed' WHERE id = {experiment_id}"
database.write(query=query)
raise Exception('Pipeline failed to run experiment')
return experiment_id
def create_project(name: str) -> int:
query = f"INSERT INTO project (name) VALUES ('{name}')"
database = Database()
return database.write(query=query)