Bug fixes Fix nondeterministic type validation error messages in DataModel.assert_isinstance (#1693, thanks @marcus-campbell !) Fix cache cleanup evicting files currently served to another process (e.g. sibling DataLoader workers) by protecting cache hits with advisory file locks (#1737) Fix shell command argument escaping in the autoscaler cloud driver to prevent syntax errors and injections (#17…
Bug fixes and improvements Add method to allow deleting data entries in hyperdatasets Add functionality for setting labels on dataviews Add hyperdataset metadata management (get/set methods) Add methods to manage mapping rules Add hyperdataset method to get an iterator for it Fix dataview race conditions and add hyperdataset method wrappers
Bug fixes and improvements Add in-memory data streaming capability to storage manager Add DataView class get method to retrieve them from the API server Add upper limit of 100 Mb for registration request payload size of data entries
Bug fixes and improvements Improve HPO resilience by pruning trials in (#1652) Deprecate parent_ids argument (but keep them usable) in favor of parent_id (#1655) Use usedforsecurity=False in hashlibeverywhere where hashing is used (see issue #1656, thanks @a-gardner1 !) Repair syntax error when instantiating a hyperdataset with a parent (#1657)
Bug fixes and improvements Add method to create a hyperdataset version snapshot Improvements in formatting via f-strings instead of .format Remove dead code about azure.storage.blob.BlobServiceClient Implement static route validator RouterService class Fix syntax-related error in the Dataset.get method (#1648, thanks to Fra from our community Slack!)
Bug fixes and improvements Add path traversal check in Dataset._merge_datasets Add "status" attribute in Dataset.list_datasets dataset return values Add hyperdataset tagging features Add support for DefaultAzureCredential Fix issue that users restricted to a project cannot create dataviews Add ability to publish a hyperdatset via single method call
Bug fixes and improvements Add optional (opt-out) parameter to prevent processing of pickled artifacts (#1619, #1621) use task.artifacts[0].get(block_unsafe_artifacts=True), or via configuration with the environment variable CLEARML_BLOCK_PICKLED_ARTIFACTS=1 via sdk.storage.block_pickled_artifacts: true Prevent resource leak during interrupted queries in GPUStatCollection._new_query_nvidia (#1617)…
Bug fixes and improvements Add default shell binary option for script execution (#1586) Add option to set plots upload destination (#1587) Add exc_info param to improve error logging (#1588, thanks @HJJ256 for raising this issue #1550 !) Raise duplicate InputModel import log from debug to info (#1582, thanks @kei-yamazaki !) Add pickle file integrity hash verification for pd.DataFrame pickled arti…
Bug fixes and improvements Add task method to remove tags Fix import clearml regression in offline mode (#1517)
Bug fixes and improvements Correct typos (thanks @thecaptain789 !) Fix syntax error on python 3.14 (#1521, thanks @mscheltienne !) Add support for boto3 S3 specific configuration (#1516, thanks @sunqing05 !) Add option to close squashed dataset (#1513, thanks @HJJ256 !) Move None check earlier in CacheContext.get_local_copy (#1570, thanks @JesseLivezey !) Work in progress in dropping Python 2 supp…
New Features and Bug Fixes Fix GPU reporting for NVIDIA_VISIBLE_DEVICES=void (#1508) Fix default example parameters for sklearn joblib Add support for ClearML Apps Gateway static routes in Gradio binding
Bug Fixes Fix broken ArgParser integration with SUPPRESS
New Features and Bug Fixes Fix space is missing from the safe characters list when quoting downloaded file names Add support for sdk.storage.http.legacy_fileservers to allow downloading data from legacy fileservers Add Python 3.14 support