- Track artifacts on S3 compatible storage (#1053)
- Update jsonschema requirement with explicit
format
specifier (#1010) - Escape inputs to SQL in Artifact LocalFileHashStorage (#1034)
- More consistent and strict way of git repository, source files and entrypoint detection (#1007)
- Moved neptune and neptune_cli to src dir (#1027)
fetch_runs_table(...)
,fetch_models_table(...)
andfetch_model_versions_table(...)
now queries only non-trashed (#1033)get_last_run
,get_run_url
,get_project
andneptune.init
marked as deprecated (#1011)- Deprecated implicit casting of objects to strings with
log
andassign
operations (#1028) - Internally extracted legacy client to
legacy
submodule (#1039)
- Management docstring adjustments (#1016)
- Few minor fixes
- Fixed
with_id
deprecation message (#1002) - Fix passing None as deprecated parameter to deprecated_parameter decorator (#1001)
- Exposed integrations related utils (#983)
- Add new with_id parameter to init functions (#985)
- Introduce filtering columns when fetching run, model and model_version tables (#986)
- Stop hanging indefinitely on wait when async data synchronization process is dead (#909)
- Finish stop() faster when async data synchronization process dies (#909)
- Added
NEPTUNE_MODE
environment variable (#928) - Added support of Service account management (#927)
- More informational exception due to plotly and matplotlib incompatibility (#960)
- Dedicated exceptions for collision and validation errors in
create_project()
(#965) - Project key is now optional in API. If it is not provided by user it is generated. (#946)
- Former
ProjectNameCollision
exception renamed to AmbiguousProjectName (#965)
-
Force jsonschema version < 4.0.0 (#922)
-
Rename and copy update for UnsupportedClientVersion and DeprecatedClientLibraryVersion (#917)
- Added fetching Models method to Project (#916)
- Fix computing of a multipart upload chunk size (#897)
- Matching all listed tags instead of any when calling
fetch_runs_table
(#899) - Fix invalid processing of delete followed by file upload in a single batch (#880)
click.echo
replaced withlogging
(#903)
- Sync only offline runs inside '.neptune' directory CLI flag (#894)
- Fix handling of server errors (#896)
- Print metadata url on stop (#883)
- Fix handling Internal Server Error (#885)
- restart upload when file changes during (#877)
- Added support for workspace visibility in Management API (#843)
- Exposed container with a property of Handler (#864)
- Restore version in neptune.new (#860)
- Methods for creating and manipulating Model Registry objects (#794)
- Renamed --run parameter to --object in
neptune sync
(previous kept as deprecated, #849) - More helpful error message on SSL validation problem (#853)
- Added names to daemon worker threads (#851)
- Stopped forwarding every attribute from Handler to Attribute (#815)
- Stripping whitespaces from Neptune API Token (#825)
- Raise proper exception when invalid token were provided (#825)
- Make status error-handling in legacy client consistent with neptune.new (#829)
- Use new file upload API (#789)
- Fixed listing available workspaces when invalid name was provided (#818)
- Added proper docstrings for Project-Level Metadata (#812)
- Fixed backward compatibility when syncing old offline data (#810)
- Prevent original numpy array from modifying (#821)
- Unpin
jsonschema<4
, pinswagger-spec-validator>=2.7.4
until bravado releases new version (#820)
- Interacting with project-level metadata (#758)
- Copy feature for non-file single value attributes (#768)
- Fix verifying data size limits in String Atoms and File.from_content (#784)
- Restore RunMode for backward compatibility (#775)
- Restore imports for backward compatibility (#777)
- Limit number of Series elements sent in single request (#780)
- Fix issue that prevented waiting for subprocesses to finish after receiving stop signal from backend (#774)
Timeout now overridable using environment var
NEPTUNE_SUBPROCESS_KILL_TIMEOUT
- Fixed multithreading bug with StdStreamCaptureLogger (#762)
- Fixed fetching numeric values in debug mode (#745)
- Ensure StdStreamCaptureLogger doesn't log after .close() (#759)
- PyTorchLightning integration is imported directly from
pytorch-lightnig
repo (#673)
- Fix issue with file upload retry buffer causing 400 bad requests (#743)
- Provide names of existing run attributes to IPython's suggestion mechanism (#740)
- Add docstrings for project management API (#738)
- Update MemberRoles to match values in the UI (#738)
- Support Artifacts in fetch_runs_table() (#728)
- Human-readable objects representation via
__repr__
(#717) - Added project management API (#695, #720)
- Performance improvements when creating several runs (#695)
- Temporarily pin
jsonschema<4
(4.0.0 is incompatible withbravado
; #719)
- API for Artifacts (#703)
- Fix exception type raised on calling missing method on Handler (#693)
- Fix leaks of descriptors
- Fix possible deadlock on synchronisation in async mode
- Fixed url building in Windows (#672)
- Fixed slashes in file operations url concatenation (#666)
- Only print info if exception actually occurred when using Run as context manager (#650)
- Added long description for PyPI (#642)
- Fixed GitPython importing during package preparation (#647)
- Checking current working directory in addition to entrypoint when looking for git repository (#633)
- Added support for Kedro integration (#641)
- Added NEPTUNE_MONITORING_NAMEPSACE environment variable (#623)
- Delete namespace (and all child fields and namespaces) (#619)
- .pop() works invoked on a field (#617)
- Logging version when using python logger integration (#622)
- Return path from requested prefix instead of root when fetching namespace (#609)
- Heuristics to help users find out they're writing legacy code with new client API or vice versa (#607)
- Lookup for projects without workspace specification and listing user projects and workspaces (#615)
- Mechanism to prevent using legacy Experiments in new-API integrations (#611)
- Prevent logging into stopped runs (#602)
- Added more informal exception for invalid API token (#601)
- Legacy client Improved stability by adding retry on failure when uploading (#604)
- Check get_ipython() for None (#598)
- Remind user about stopping runs in interactive console and notebooksz (#595)
- Updating error messages and links to docs (#593)
- Added support for fast.ai integration (#590)
- Allow for updating an already assigned Namespace instead of failing with errors (mostly affects Optuna integration) (#585)