DeepLake

This release introduces support for binary data within dataset and column metadata and removes the dependency on libatomic.

Version 4.5.1 focuses heavily on performance improvements through memory allocation changes (mimalloc, simdjson) and introduces significant new features like full null value support and dataset statistics gathering.

This release includes improved support for NULL values in Deeplake and resolves a concurrency issue during row and column creation.

This release focuses on infrastructure improvements, including better build integration via cmake/pkg-config for the Deeplake API and significant enhancements to storage access performance and PostgreSQL batch ingestion.

No release notes provided.

This release introduces significant storage and API enhancements, including a new list_dirs API and mesh type support, alongside substantial performance improvements in cache operations and bug fixes across core functionality and indexing.

This release introduces significant indexing improvements, including JSON data indexing support and better progress indication. It also resolves several critical bugs related to compilation, Windows testing, and browser stability.

This release introduces support for the link to bytes data type and improves text column flexibility. Several bugs related to error handling and PNG linking have also been resolved.

This release introduces PostgreSQL 18 compatibility and significant performance optimizations, including adaptive backoff and cgroup-aware concurrency, alongside various bug fixes.

This release introduces significant enhancements to pg_deeplake, including new data type support and performance improvements, alongside architectural refactoring and bug fixes.

This release primarily updates dependency compatibility by allowing usage with numpy version 2.

Deeplake 4.3.0 is a major update introducing comprehensive support for video data, enhanced indexing for numeric types, and significant improvements to CSV import/export functionality.

This minor release (3.9.51) primarily addresses a bug in the dependency resolver.

This release includes an update to version 3.9.50 and refactors the frame extraction logic along with improvements to frame rate retrieval.

This patch release introduces a base mmsegmentation dataset class to facilitate integration with mmseg 1.x, requiring users to inherit from it.

This release introduces the new Audio type and fixes a conversion bug related to the Polygon type from V3.

This release introduces several new features including autocommit, TQL enhancements (AVG function, improved SAMPLE BY), and expanded support for data types and integrations like zlib compression for SegmentMask and float16/bfloat16 indexing.

This release introduces new data loading capabilities with `from_csv` and byte support for `from_parquet`, alongside new compression options for `SegmentMask`.

This release introduces a new sampler setter function and implements a hierarchical namespace structure.

This release introduces significant enhancements to indexing capabilities, batch query support, and rich type handling within Structs, alongside important bug fixes for data consistency and chunking.

This release introduces data file compaction, improved image handling, a new text comparison index type, and enhanced asynchronous operation support, including async iteration over batches.

Deeplake 4.2 introduces automatic commit compaction for faster opens and stabilizes the asynchronous API, alongside metadata copying in `deeplake.like`.

This minor release addresses a specific concurrency issue related to dataset size tracking during row deletion.

This release introduces improvements in version control, adds support for float16 and bfloat16 data types, and integrates OpenTelemetry for enhanced observability.

This release updates internal versions and introduces support for Labelbox IAM integration.

This major release introduces dataset branching, enhanced query capabilities including direct credential passing and full JOIN support, and improved Labelbox integration.

This release focuses on stability and correctness, primarily fixing bugs related to S3 redirects, virtual column dtypes, iteration handling, and network error reporting.

This release primarily focuses on fixing a visualization issue for local datasets.

This release introduces an exception for empty class names and resolves a palette issue. Version updates were also applied for the 3.9.42 release.

This release fixes a high memory usage issue during json column queries and introduces the ability to append arrays of nulls to image type columns.

This release focuses on performance improvements for dataset loading and querying, introduces native support for MMR search, and adds the EmbeddingMatrix index type.

This release primarily updates internal dependencies to version 3.9.41 and removes the unnecessary `labelbox_client` argument from labelbox integration functions.

This release integrates the mmseg 1.x version and updates internal version dependencies.

This release includes refactoring for the Labelbox integration, updates to internal dependencies, and improved resilience for Azure storage authentication.

This release primarily updates internal versions for the 3.9.38 release and includes a fix for an issue concerning max_view.

This release introduces support for new data types including dicom, nifti, and point, alongside performance improvements in index generation and new search capabilities.

This release includes minor updates, specifically updating versions for the 3.9.37 release and updating the client after a credentials exchange.

This release includes minor updates, specifically updating versions for the 3.9.36 release and updating the client after a credentials exchange.

This release introduces support for the Point type, adds automation for COCO-like dataset ingestion via from_coco, and includes several bug fixes related to query performance and data handling.