Deploying training projects on Dataiku DSS

dataiku_dss_training_deployer.py

# Parameters
number_of_groups = 5
number_of_users_per_group = 3
generic_password = "password4Training!!"
analysis_id, mltask_id = u'bOkH96fi', u'E7bXC2fR'
saved_model_id="sXGr0kU2"


# Group creation
import dataiku
client = dataiku.api_client()

group_name_list = [i["name"] for i in client.list_groups()]
target_group_definition = {
 u'admin': False,
 u'mayCreateAuthenticatedConnections': True,
 u'mayCreateCodeEnvs': True,
 u'mayCreateProjects': True,
 u'mayDevelopPlugins': True,
 u'mayEditLibFolders': True,
 u'mayManageCodeEnvs': True,
 u'mayManageUDM': False,
 u'mayViewIndexedHiveConnections': True,
 u'mayWriteSafeCode': True,
 u'mayWriteUnsafeCode': True,
}
for i in range(number_of_groups):
    group_name = "group"+str(i+1)
    group = client.get_group(group_name)
    if group_name in group_name_list:
        group.delete()
    client.create_group(group_name, description="Training "+group_name, source_type='LOCAL')    
    group_definition = group.get_definition()
    for k,v in target_group_definition.items():
        group_definition[k] = v
    group.set_definition(group_definition)

# User group assignment

list_user_login = [i["login"] for i in client.list_users()]
for i in range(number_of_groups*number_of_users_per_group):
    user_login = "user"+str(i+1)
    user = client.get_user(user_login)
    group_assignment = "group"+str((i // number_of_users_per_group) + 1)
    if user_login in list_user_login:
        user.delete()
    client.create_user(user_login, 
                       generic_password, 
                       display_name=user_login, 
                       source_type='LOCAL', 
                       groups=[group_assignment],
                       profile='DATA_SCIENTIST')

# Project cloning

print(client.list_project_keys())

current_project_key = dataiku.default_project_key()
current_project = client.get_project(current_project_key)
for i in range(number_of_groups):
    project_key = current_project_key+"_"+str(i+1)
    project = client.get_project(project_key)
    group_name = "group"+str(i+1)    
    if project_key in client.list_project_keys():
        project.delete()
    export_stream = current_project.get_export_stream(options={
        "exportAllDatasets":  True,
        "exportSavedModels" : True
    })
    with export_stream as s:
        client.prepare_project_import(s).execute(settings={"targetProjectKey":project_key})
    target_permissions = {
        u'owner': u'dataiku',
        u'permissions': [
           {
                u'admin': False,
                u'exportDatasetsData': True,
                u'group': group_name,
                u'manageAdditionalDashboardUsers': True,
                u'manageDashboardAuthorizations': True,
                u'manageExposedElements': True,
                u'moderateDashboards': True,
                u'readDashboards': True,
                u'readProjectContent': True,
                u'runScenarios': False,
                u'writeDashboards': True,
                u'writeProjectContent': True
            }
       ]
    }
    project.set_permissions(target_permissions)

# Model redeployment

print(current_project.list_ml_tasks())

for i in range(number_of_groups):
    project_key = current_project_key+"_"+str(i+1)
    project = client.get_project(project_key)
    mltask = project.get_ml_task(analysis_id, mltask_id)
    mltask.start_train()
    mltask.wait_train_complete()
    print(mltask.get_trained_models_ids())
    model_id = mltask.get_trained_models_ids()[0]
    mltask.redeploy_to_flow(model_id, saved_model_id=saved_model_id)