You can connect to Neptune using one of the five connection modes:
You can select mode by providing
mode parameter to the
import neptune.new as neptune# A default connection mode is the asynchronous mode# Other possible values are "async", "sync", "offline", "debug"CONNECTION_MODE = "async"run = neptune.init(name="My new run", mode=CONNECTION_MODE)
All tracking calls (like
save()) are non-blocking. The tracked data is temporarily stored on the local disk and synchronized with the Neptune server in the background by a separate synchronization thread.
We recommend using Neptune with persistent disks to be able to restore tracked data in case your machine restarts (e.g. spot instances) or in case of connectivity issues with Neptune servers.
Neptune triggers disk flushing:
Every 5 seconds (this period is configurable, see
When you invoke
At the end of the run (destruction of the
Tracking calls in asynchronous mode do not throw exceptions related to connectivity or metadata consistency issues (more in the distributed computing section). Issues related to connectivity or metadata consistency issues will be printed to
When you use asynchronous (default) connection mode and there is a problem with the connection to Neptune servers (e.g. caused by Internet connectivity issues), the Neptune Client Library will continue to locally track your metadata and will continuously try to re-establish connection with Neptune servers.
Once your run is finished, if the connection has not been reestablished, Neptune will continue trying for another 5 minutes. After this period the process will end. All the unsuccessful tracking calls are stored on the local disk. The unsynchronized metadata from the local disk can be uploaded later via Neptune CLI (Command Line Interface)
Tracking calls return only after the Neptune server responds that the data was stored.
import neptune.new as neptunerun = neptune.init(name="My new run", mode="sync")
In this mode, no connection to Neptune servers is established. Instead, all the tracked metadata is stored on the local disk and can be uploaded to Neptune servers manually via
neptune sync command.
import neptune.new as neptunerun = neptune.init(name="My new run", mode="offline")
Whether you experience connectivity issues or you are working in offline mode your data is stored safely locally. You can use Neptune CLI (Command Line Interface) to check the synchronization status and synchronize data with Neptune servers.
You can list unsynchronized runs by using
# List unsynchronized runs in the current directoryneptune status# List unsynchronized runs in the given pathneptune status --path PATH_TO_DIRECTORY# Access status command help and examplesneptune status --help
Synchronize local data with Neptune servers with
# Synchronize all runs in the current directoryneptune sync# Synchronize all runs in the given pathneptune sync --path PATH_TO_DIRECTORY# Synchronize only runs NPT-42 and NPT-43neptune sync -run workspace/project/NPT-42 -run workspace/project/NPT-43# Synchronise all runs in the current directory# sending offline runs to project "workspace/project"neptune sync -p workspace/project# Synchronize the offline run a1561719-b425-4000-a65a-b5efb044d6bb# to project "workspace/project"neptune sync -p workspace/project -run offline/a1561719-b425-4000-a65a-b5efb044d6bb# Access sync command help and examplespython -m neptune.new.cli sync --help
A read-only mode is useful when you need only to fetch your run's metadata. For example, you are connecting to Neptune to perform a specific analysis on your run's metadata that is not possible inside Neptune. In this mode:
No write operation is possible. Any write-like operation will be ignored.
No metadata is sent to Neptune. In particular stdout and stderr are not captured, and no hardware metrics are being sent.
The run never switches to Active status, because of a read-only connection.
import neptune.new as neptune# Connecting to run wih SUN-2 identifierrun = neptune.init(run="SUN-2", mode="read-only")
Debug mode can come in handy when you are debugging your code and would like to not pollute your project. In this mode, no calls are made to Neptune servers, regardless of what happens in the code. In contrast to Offline mode, all data are stored only in memory.
import neptune.new as neptunerun = neptune.init(name="My new run", mode="debug")