Admin API¶

This page describes the admin API, which lets you manage administrative operations in the Driverless AI Python client.

Initialize a client with your server credentials and store them in the variable client.

In [4]:

import driverlessai
client = driverlessai.Client(address='http://localhost:12345', username="py", password="py")

import driverlessai client = driverlessai.Client(address='http://localhost:12345', username="py", password="py")

You can check whether the connected user has admin privileges as follows.

In [13]:

client.admin.is_admin

client.admin.is_admin

Out[13]:

True

Loading all users¶

You can list all the users of Driverless AI as follows.

In [43]:

client.admin.list_users()

client.admin.list_users()

Out[43]:

['alice', 'bob', 'charlie']

Loading all currently logged in users¶

All of the logged in users of a Driverless AI server can be listed as follows. Note that this functionality is only supported for Driverless AI server version 1.10.5 and later.

In [44]:

client.admin.list_current_users()

client.admin.list_current_users()

Out[44]:

['alice', 'bob']

List the datasets of a specified user¶

You can list the datasets created by a specific user as follows.

In [38]:

datasets = client.admin.list_datasets(username="alice")

datasets = client.admin.list_datasets(username="alice")

This returns a list of objects of DatasetProxy and it has the following properties.

In [45]:

dataset = datasets[0]
print("Dataset key : " , dataset.key)
print("Dataset name : " , dataset.name)
print("Dataset owner : " , dataset.owner)

dataset = datasets[0] print("Dataset key : " , dataset.key) print("Dataset name : " , dataset.name) print("Dataset owner : " , dataset.owner)

Dataset key :  02fdc4d4-cc7d-11ed-ae20-ac1f6b6b4a0e
Dataset name :  retail-data.csv
Dataset owner :  alice

You can also delete a dataset as follows.

In [8]:

dataset.delete()

dataset.delete()

List the experiments of a specified user¶

You can list all the experiments created by a specific user as follows.

In [41]:

experiments = client.admin.list_experiments(username="alice")

experiments = client.admin.list_experiments(username="alice")

This returns a list of objects of ExperimentProxy and it provides the following methods.

In [42]:

experiment = experiments[0]
print("Completed : " , experiment.is_complete())
print("Running : " , experiment.is_running())
print("Experiment status : " , experiment.status(verbose=1))

experiment = experiments[0] print("Completed : " , experiment.is_complete()) print("Running : " , experiment.is_running()) print("Experiment status : " , experiment.status(verbose=1))

Completed :  True
Running :  False
Experiment status :  Complete 100.00%

Transfer data between users¶

Here, all the data, including datasets and experiments, is transferred to the other user.

In [29]:

client.admin.transfer_data(from_user="alice", to_user="bob")

client.admin.transfer_data(from_user="alice", to_user="bob")

You can now view the datasets created by Alice under Bob's datasets.

In [30]:

client.admin.list_datasets(username="bob")

client.admin.list_datasets(username="bob")

Out[30]:

[<driverlessai._admin.DatasetProxy at 0x12ac98190>,
 <driverlessai._admin.DatasetProxy at 0x12ac989a0>]

List scheduled experiments in the DAI server and manipulate execution priorities¶

Here, you have the ability to list the scheduled experiments for a specific user on the DAI server.

In [11]:

scheduled_exps = client.admin.list_scheduled_experiments(user="alice")

scheduled_exps = client.admin.list_scheduled_experiments(user="alice")

/tmp/ipykernel_72777/528868256.py:1: UserWarning: 'Admin.list_scheduled_experiments' is a beta API that is subject to future changes.
  scheduled_exps = client.admin.list_scheduled_experiments(user="alice")

This API returns a list of ScheduledTask objects, each containing the following properties.

In [12]:

print("Task ID: ", scheduled_exps[5].key)
print("Driverless AI procedure: ", scheduled_exps[5].procedure)
print("Entity ID: ", scheduled_exps[5].entity_id)
print("Task queue: ", scheduled_exps[5].queue)
print("Timeout: ", scheduled_exps[5].timeout)
print("Username: ", scheduled_exps[5].user)
print("Queueing time: ", scheduled_exps[5].queueing_time)

print("Task ID: ", scheduled_exps[5].key) print("Driverless AI procedure: ", scheduled_exps[5].procedure) print("Entity ID: ", scheduled_exps[5].entity_id) print("Task queue: ", scheduled_exps[5].queue) print("Timeout: ", scheduled_exps[5].timeout) print("Username: ", scheduled_exps[5].user) print("Queueing time: ", scheduled_exps[5].queueing_time)

Task ID:  08044682-b12e-11ef-b7d9-ac1f6b00be4b
Driverless AI procedure:  auto_dl_easy_api_wrapper
Entity ID:  07f475a4-b12e-11ef-b7d9-ac1f6b00be4b
Task queue:  p:gpu
Timeout:  None
Username:  alice
Queueing time:  1733199673

You are able to modify the execution priority of a specific experiment.

In [13]:

scheduled_exps[5].update_schedule_priority(priority=1)

scheduled_exps[5].update_schedule_priority(priority=1)

As a result, this experiment will now be assigned the highest priority in the priority queue.

In [14]:

new_list = client.admin.list_scheduled_experiments(user="alice")
print("1st on GPU: ", new_list[0].entity_id)

new_list = client.admin.list_scheduled_experiments(user="alice") print("1st on GPU: ", new_list[0].entity_id)

/tmp/ipykernel_72777/4107197592.py:1: UserWarning: 'Admin.list_scheduled_experiments' is a beta API that is subject to future changes.
  new_list = client.admin.list_scheduled_experiments(user="alice")

1st on GPU:  07f475a4-b12e-11ef-b7d9-ac1f6b00be4b

Additionally, you have the capability to reorder the execution priorities of multiple scheduled experiments.

In [ ]:

client.admin.reorder_experiment_schedule(
    [
        {
            "username": new_list[1].user,
            "key": new_list[1].entity_id
        },
        {
            "username": new_list[0].user,
            "key": new_list[0].entity_id
        }
    ]
)

client.admin.reorder_experiment_schedule( [ { "username": new_list[1].user, "key": new_list[1].entity_id }, { "username": new_list[0].user, "key": new_list[0].entity_id } ] )