def test_download_and_save_from_mlflow(self, tmp_path):
saving_path = Path(tmp_path, "file.txt")
with requests_mock.Mocker() as mocked_req:
mocked_req.get("http://test-tracking-uri:8080/get-artifact?path=full_model.pkl&run_uuid=123-abc",
text="Hello World")
ModelCache.download_and_save_from_ml_flow(saving_path, "123-abc")
assert saving_path.read_text() == "Hello World"
def test_get_latest_model_no_latest(self, tmp_path):
cache = ModelCache(tmp_path)
def get_search_return_values(experiment_ids, *args, **kwargs):
df = pd.DataFrame(columns=list(cache.columns_of_interest.keys()), data=[{
'run_id': "123",
'tags.mlflow.runName': '1',
'end_time': datetime(2020, 8, 29, 8, 0, 0),
'params.MLPipelineParamsName': 'default',
'params.FeatureSetName': 'default',
'params.AlgorithmName': 'default',
'params.AlgorithmParamsName': 'default',
'tags.DidPassAcceptanceTest': 'no'
}])
df["end_time"] = pd.to_datetime(df["end_time"])
return df
def get_experiment_id(scenario, *args, **kwargs):
if scenario == "houses":
return None
else:
return SimpleNamespace(**{'experiment_id': "id_" + scenario})
model_as_bytes = self.get_sample_model_path().read_bytes()
with requests_mock.Mocker() as mocked_req:
mocked_req.get("http://test-tracking-uri:8080/get-artifact?path=full_model.pkl&run_uuid=123",
content=model_as_bytes)
mlflow.get_experiment_by_name = MagicMock(side_effect=get_experiment_id)
mlflow.search_runs = MagicMock(side_effect=get_search_return_values)
loaded_model = cache.get_loaded_model_for_scenario_and_run_id("groceries", "latest")
assert loaded_model is None
def test_return_a_couple_models(self):
cache = ModelCache()
def get_search_return_values(experiment_ids, *args, **kwargs):
df = pd.DataFrame(columns=list(cache.columns_of_interest.keys()),
data=[{
'run_id': "123",
'tags.mlflow.runName': 'my_run',
'tags.mlflow.BuildNumber': '4',
'end_time': datetime(2020, 8, 29, 8, 0, 0),
'params.MLPipelineParamsName': 'default',
'params.FeatureSetName': 'default',
'params.AlgorithmName': 'default',
'params.AlgorithmParamsName': 'default',
'tags.DidPassAcceptanceTest': 'no'
}, {
'run_id': "456",
'tags.mlflow.runName': 'my_second_run',
'tags.mlflow.BuildNumber': '5',
'end_time': datetime(2020, 8, 29, 9, 0, 0),
'params.MLPipelineParamsName': 'default',
'params.FeatureSetName': 'default',
'params.AlgorithmName': 'default',
'params.AlgorithmParamsName': 'default',
'tags.DidPassAcceptanceTest': 'yes'
}])
df["end_time"] = pd.to_datetime(df["end_time"])
return df
def get_experiment_id(scenario, *args, **kwargs):
if scenario == "houses":
return None
else:
return SimpleNamespace(**{'experiment_id': "id_" + scenario})
mlflow.get_experiment_by_name = MagicMock(
side_effect=get_experiment_id)
mlflow.search_runs = MagicMock(side_effect=get_search_return_values)
available_models = cache.list_available_models_from_ml_flow()
assert set(available_models.keys()) == {"groceries", "iris"}
assert len(available_models["groceries"]) == 2
assert [x["run_id"]
for x in available_models["groceries"]] == ['123', '456']
assert [
x["is_latest"] for x in available_models["groceries"]
if x["run_id"] == "456"
][0]
assert not [
x["is_latest"]
for x in available_models["groceries"] if x["run_id"] == "123"
][0]
def test_list_no_models_available(self):
cache = ModelCache()
def get_search_return_values(experiment_ids, *args, **kwargs):
df = pd.DataFrame(columns=list(cache.columns_of_interest.keys()))
df["end_time"] = pd.to_datetime(df["end_time"])
return df
def get_experiment_id(scenario, *args, **kwargs):
return SimpleNamespace(**{'experiment_id': "id_" + scenario})
mlflow.get_experiment_by_name = MagicMock(side_effect=get_experiment_id)
mlflow.search_runs = MagicMock(side_effect=get_search_return_values)
available_models = cache.list_available_models_from_ml_flow()
assert available_models == {"groceries": [], "houses": []}
def test_is_not_latest_deployable_model(self):
row = {
"ml_pipeline_params_name": 'default',
"feature_set_name": 'default',
"algorithm_name": 'my_params',
"algorithm_params_name": 'default',
"passed_acceptance_test": 'yes'
}
assert not ModelCache.is_latest_deployable_model(row)
"""
Web app
"""
import logging
import requests
from flask import Flask, request, render_template
from cd4ml.logger.fluentd_logging import FluentdLogger
from cd4ml.webapp.model_cache import ModelCache
from cd4ml.webapp.webapp_data_scorer import get_form_from_model
logger = logging.getLogger(__name__)
fluentd_logger = FluentdLogger()
cache = ModelCache()
ERROR_NO_AVAILABLE_LATEST_MODEL = "No model is available for the latest scenario {}. " \
"Please re-run jenkins and reload this page"
ERROR_NO_MODEL_AT_LOCATION = "No model is available for this identifier scenario {}. " \
"Please navigate back to the model picker and try again."
app = Flask(__name__,
template_folder='webapp/templates',
static_folder='webapp/static')
def make_page_for_scenario_and_identifier(scenario_name, identifier, request_data):
model = cache.get_loaded_model_for_scenario_and_run_id(scenario_name, identifier)
header_text, form_div, prediction = get_form_from_model(scenario_name, identifier, model,
initial_values=request_data)
return header_text, form_div, prediction