3LC Python Package Version 2.5

2.5.1

Enhancements and Fixes

  • Fixed an issue with version reported by 3lc --version

2.5.0

With the release of the 2.5.0 version of the 3LC Python Package, we have changed our development processes to allow for incremental semantic versioning. This is important as users accumulate dependencies on the public API that we want to ensure do not break unexpectedly. Going forward, we will also have incrementally numbered patch releases of the 3lc wheel on PyPI (2.5.0, then 2.5.1, and so on).

Features

  • [12427] Updated processes to support proper incremental semantic versioning for the 3LC Python package

Enhancements and Fixes

  • [11967] Clean up the tlc package namespaces, removing everything not created by the tlc package itself, so that e.g. types from torch are not brought into the tlc namespace

  • [12523] Updated error message to be more informative when importing tlc without torch or torchvision installed

  • [11614] Made it so SampleType.from_schema infers sample type from ScalarValue types

  • [12506] Fixed a bug where where tables whose if_exists option was set to "rename" were not persisted when name collision happened

  • [12400] Always create row cache for TableFromYolo, which also avoids parsing the YAML file and scanning for images/labels several times, which can be time-consuming

  • [12501] Fixed a bug where TableFromHuggingFace would crash if the feature description contained a list. This list will now be interpreted in the same way as the Sequence feature.

  • [12425] Changed the time between starting indexing scans from 200ms to 1s to reduce load on CPU

  • [12296] Implemented a couple of small optimizations for indexing scans

  • [12353] Fixed bug where incorrect help message was printed for tlccli

  • [12524] Wrap lines in the log widget in the object service TUI

  • [12364] Made fixes to telemetry reporting to report the correct environment, to include more useful information, and to send data on normal shutdown

  • [12344] Added joblib dependency as part of umap and pacmap extras

  • [12441] Changed example notebooks to use PaCMAP instead of UMAP for embeddings

Known Issues

  • The tlc Python package does not detect, handle, or support NaN (Not-a-Number) values in tlc.Table, and their presence may lead to unpredictable behavior or inconsistencies within the system.