Frequently Asked Questions

What Driverless AI server version does the client work with?

The client version number indicates the most recent Driverless AI server supported by that specific client version. However, all client versions are backwards compatible with Driverless AI servers down to version 1.8.0. See the Supported Driverless AI Servers section in the overview page for the specific Driverless AI server versions supported.

How can I avoid hardcoding my password?

You can use the getpass module from the Python standard library to prompt the user for a password:

[1]:
import driverlessai
import getpass

dai = driverlessai.Client(
    address='http://localhost:12345',
    username='py',
    password=getpass.getpass("Enter Driverless AI password: ")
)
Enter Driverless AI password: ········

How can I suppress InsecureRequestWarning messages when initializing the client with verify=False?

While it’s not recommended to connect to an unverified server, if you understand the risks you can suppress InsecureRequestWarning messages with:

[1]:
import requests
from urllib3.exceptions import InsecureRequestWarning

requests.packages.urllib3.disable_warnings(category=InsecureRequestWarning)

Why is verify=false ignored when initializing the client ?

Please unset the REQUESTS_CA_BUNDLE environment variable if it exists. See https://github.com/psf/requests/issues/3829 for more information.

Can I download to an external file system, such as s3?

Yes! Any file system package that implements FSSPEC should be able to create a file system object that can be passed to any download method in driverlessai. See FSSPEC for a list of known implementations. For example, we can use s3fs to download experiment artifacts to a s3 bucket named h2oai:

[ ]:
import s3fs

ex = dai.experiments.get("<UUID>")

fs = s3fs.S3FileSystem(anon=False)

s3_path = ex.artifacts.download(
    dst_dir="h2oai/data/<UUID>/",
    file_system=fs
)