def verify(run: kfp_server_api.ApiRun, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
verify_tasks(t, tasks)
def verify(run: kfp_server_api.ApiRun, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
pprint(tasks)
output_artifact = tasks['output-artifact']
output = [
a for a in output_artifact.outputs.artifacts if a.name == 'Output'
][0]
pprint(output)
host = os.environ['MINIO_SERVICE_SERVICE_HOST']
port = os.environ['MINIO_SERVICE_SERVICE_PORT']
minio = Minio(f'{host}:{port}',
access_key='minio',
secret_key='minio123',
secure=False)
bucket, key = output.uri[len('minio://'):].split('/', 1)
print(f'bucket={bucket} key={key}')
response = minio.get_object(bucket, key)
data = response.read().decode('UTF-8')
t.assertEqual(data, 'firstsecond\nfirstsecond\nfirstsecond')
def verify_with_specific_pipeline_root(run: kfp_server_api.ApiRun,
mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
verify_tasks(t, tasks)
verify_artifacts(t, tasks, 'minio://mlpipeline/override/artifacts')
def verify(run, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
t.assertEqual(17, tasks['add-3'].outputs.parameters['sum'],
'add result should be 17')
def verify(run: kfp_server_api.ApiRun, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
pprint(tasks)
roc_visualization = tasks['roc-visualization']
output = [
a for a in roc_visualization.outputs.artifacts
if a.name == 'mlpipeline_ui_metadata'
][0]
pprint(output)
t.assertEqual(
roc_visualization.get_dict()['outputs']['artifacts'][0]['name'],
'mlpipeline_ui_metadata')
def verify(mlmd_connection_config, run: kfp_server_api.ApiRun, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
# Verify MLMD state
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
task_names = [*tasks.keys()]
t.assertEqual(task_names, ['echo-msg', 'print-file', 'download-from-gcs'])
for task in task_names:
pprint(f'======= {task} =======')
pprint(tasks.get(task).get_dict())
t.assertEqual(
tasks.get('echo-msg').inputs.parameters.get('msg'),
'exit!',
)
def verify(run: kfp_server_api.ApiRun, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
# uncomment to debug
# pprint(tasks)
vis = tasks['create-tensorboard-visualization']
pprint(vis)
mlpipeline_ui_metadata = vis.outputs.artifacts[0]
t.assertEqual(mlpipeline_ui_metadata.name, 'mlpipeline-ui-metadata')
# download artifact content
storage_client = storage.Client()
blob = storage.Blob.from_string(mlpipeline_ui_metadata.uri, storage_client)
data = blob.download_as_text(storage_client)
data = json.loads(data)
t.assertTrue(data["outputs"][0]["source"])
# source is a URI that is generated differently each run
data["outputs"][0]["source"] = "<redacted>"
t.assertEqual(
{
"outputs": [{
"type":
"tensorboard",
"source":
"<redacted>",
"image":
"gcr.io/deeplearning-platform-release/tf2-cpu.2-3:latest",
"pod_template_spec": {
"spec": {
"containers": [{
"env": [{
"name": "AWS_ACCESS_KEY_ID",
"valueFrom": {
"secretKeyRef": {
"name": "mlpipeline-minio-artifact",
"key": "accesskey"
}
}
}, {
"name": "AWS_SECRET_ACCESS_KEY",
"valueFrom": {
"secretKeyRef": {
"name": "mlpipeline-minio-artifact",
"key": "secretkey"
}
}
}, {
"name": "AWS_REGION",
"value": "minio"
}, {
"name": "S3_ENDPOINT",
"value": "minio-service:9000"
}, {
"name": "S3_USE_HTTPS",
"value": "0"
}, {
"name": "S3_VERIFY_SSL",
"value": "0"
}]
}]
}
}
}]
}, data)
def verify(run: kfp_server_api.ApiRun, mlmd_connection_config, **kwargs):
t = unittest.TestCase()
t.maxDiff = None # we always want to see full diff
t.assertEqual(run.status, 'Succeeded')
client = KfpMlmdClient(mlmd_connection_config=mlmd_connection_config)
tasks = client.get_tasks(run_id=run.id)
task_names = [*tasks.keys()]
t.assertCountEqual(task_names, ['preprocess', 'train'], 'task names')
pprint(tasks)
preprocess = tasks['preprocess']
train = tasks['train']
pprint(preprocess.get_dict())
t.assertEqual(
{
'inputs': {
'parameters': {
'message': 'message',
}
},
'name': 'preprocess',
'outputs': {
'artifacts': [{
'metadata': {
'display_name': 'output_dataset_one'
},
'name': 'output_dataset_one',
'type': 'system.Dataset'
}, {
'metadata': {
'display_name': 'output_dataset_two_path'
},
'name': 'output_dataset_two_path',
'type': 'system.Dataset'
}],
'parameters': {
'output_bool_parameter_path': True,
'output_dict_parameter_path': {
"A": 1,
"B": 2
},
'output_list_parameter_path': ["a", "b", "c"],
'output_parameter_path': 'message'
}
},
'type': 'system.ContainerExecution',
'state': Execution.State.COMPLETE,
},
preprocess.get_dict(),
)
t.assertEqual(
{
'inputs': {
'artifacts': [{
'metadata': {
'display_name': 'output_dataset_one'
},
'name': 'dataset_one_path',
'type': 'system.Dataset'
}, {
'metadata': {
'display_name': 'output_dataset_two_path'
},
'name': 'dataset_two',
'type': 'system.Dataset'
}],
'parameters': {
'input_bool': True,
'input_dict': {
"A": 1.0,
"B": 2.0,
},
'input_list': ["a", "b", "c"],
'message': 'message'
}
},
'name': 'train',
'outputs': {
'artifacts': [{
'metadata': {
'display_name': 'model',
'accuracy': 0.9,
},
'name': 'model',
'type': 'system.Model'
}],
},
'type': 'system.ContainerExecution',
'state': Execution.State.COMPLETE,
},
train.get_dict(),
)
