Sourced from sentence-transformers's releases.

v5.2.0 - CrossEncoder multi-processing, multilingual NanoBEIR evaluators, similarity score in mine_hard_negatives, Transformers v5 support

This minor release introduces multi-processing for CrossEncoder (rerankers), multilingual NanoBEIR evaluators, similarity score outputs in mine_hard_negatives, Transformers v5 support, Python 3.9 deprecations, and more.

Install this version with

# Training + Inference
pip install sentence-transformers[train]==5.2.0
Inference only, use one of:
pip install sentence-transformers==5.2.0
pip install sentence-transformers[onnx-gpu]==5.2.0
pip install sentence-transformers[onnx]==5.2.0
pip install sentence-transformers[openvino]==5.2.0

CrossEncoder Multi-processing

The CrossEncoder class now supports multiprocessing for faster inference on CPU and multi-GPU setups. This brings CrossEncoder functionality in line with the existing multiprocessing capabilities of SentenceTransformer models, allowing you to use multiple CPU cores or GPUs to speed up both the predict and rank methods when processing large batches of sentence pairs.

The implementation introduces these new methods, mirroring the SentenceTransformer approach:

• start_multi_process_pool() - Initialize a pool of worker processes
• stop_multi_process_pool() - Clean up the worker pool

Usage is straightforward with the new pool parameter:

from sentence_transformers.cross_encoder import CrossEncoder
def main():
model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2')
# Start a pool of workers
pool = model.start_multi_process_pool()
Use the pool for faster inference
scores = model.predict(sentence_pairs, pool=pool)
rankings = model.rank(query, documents, pool=pool)
Clean up when done
model.stop_multi_process_pool(pool)

if name == "main":
main()

Or simply pass a list of devices to device to have predict and rank automatically create a pool behind the scenes.

from sentence_transformers.cross_encoder import CrossEncoder
</tr></table>

... (truncated)

• d4d70de Release v5.2.0
• b16248c docs: add release notes summary for v5.2 on main page (#3592)
• 2571f24 docs: update NanoBEIR collection links and descriptions for evaluators (#3591)
• 32cb5de Mine hard negatives: optionally output similarity scores (#3506)
• 9f9ad08 Extend NanoBEIR evaluators to support custom NanoBEIR datasets (#3583)
• 21dfa69 feat/deps: make Pillow an optional dependency (#3589)
• 67ff0fe Skip test_train_stsb tests; triggers rate limit too often (#3590)
• 4cb5cdc [deps] Use optimum-onnx now that both optimum-onnx and optimum-intel can us...
• 3f80d1c [feat] add multiprocessing support for Cross Encoder (#3580)
• f9724ff [feat] Allow transformers v5.0, add CI for transformers <v5 and >=v5 (#3586)
• Additional commits viewable in compare view

Sourced from black's releases.

26.1.0

Highlights

Introduces the 2026 stable style (#4892), stabilizing the following changes:

• always_one_newline_after_import: Always force one blank line after import statements, except when the line after the import is a comment or an import statement (#4489)
• fix_fmt_skip_in_one_liners: Fix # fmt: skip behavior on one-liner declarations, such as def foo(): return "mock" # fmt: skip, where previously the declaration would have been incorrectly collapsed (#4800)
• fix_module_docstring_detection: Fix module docstrings being treated as normal strings if preceded by comments (#4764)
• fix_type_expansion_split: Fix type expansions split in generic functions (#4777)
• multiline_string_handling: Make expressions involving multiline strings more compact (#1879)
• normalize_cr_newlines: Add \r style newlines to the potential newlines to normalize file newlines both from and to (#4710)
• remove_parens_around_except_types: Remove parentheses around multiple exception types in except and except* without as (#4720)
• remove_parens_from_assignment_lhs: Remove unnecessary parentheses from the left-hand side of assignments while preserving magic trailing commas and intentional multiline formatting (#4865)
• standardize_type_comments: Format type comments which have zero or more spaces between # and type: or between type: and value to # type: (value) (#4645)

The following change was not in any previous stable release:

• Regenerated the _width_table.py and added tests for the Khmer language (#4253)

This release alo bumps pathspec to v1 and fixes inconsistencies with Git's .gitignore logic (#4958). Now, files will be ignored if a pattern matches them, even if the parent directory is directly unignored. For example, Black would previously format exclude/not_this/foo.py with this .gitignore:

exclude/
!exclude/not_this/

Now, exclude/not_this/foo.py will remain ignored. To ensure exclude/not_this/ and all of it's children are included in formatting (and in Git), use this .gitignore:

*/exclude/*
!*/exclude/not_this/

This new behavior matches Git. The leading */ are only necessary if you wish to ignore matching subdirectories (like the previous behavior did), and not just matching root

... (truncated)

Sourced from black's changelog.

26.1.0

Highlights

Introduces the 2026 stable style (#4892), stabilizing the following changes:

• always_one_newline_after_import: Always force one blank line after import statements, except when the line after the import is a comment or an import statement (#4489)
• fix_fmt_skip_in_one_liners: Fix # fmt: skip behavior on one-liner declarations, such as def foo(): return "mock" # fmt: skip, where previously the declaration would have been incorrectly collapsed (#4800)
• fix_module_docstring_detection: Fix module docstrings being treated as normal strings if preceded by comments (#4764)
• fix_type_expansion_split: Fix type expansions split in generic functions (#4777)
• multiline_string_handling: Make expressions involving multiline strings more compact (#1879)
• normalize_cr_newlines: Add \r style newlines to the potential newlines to normalize file newlines both from and to (#4710)
• remove_parens_around_except_types: Remove parentheses around multiple exception types in except and except* without as (#4720)
• remove_parens_from_assignment_lhs: Remove unnecessary parentheses from the left-hand side of assignments while preserving magic trailing commas and intentional multiline formatting (#4865)
• standardize_type_comments: Format type comments which have zero or more spaces between # and type: or between type: and value to # type: (value) (#4645)

The following change was not in any previous stable release:

• Regenerated the _width_table.py and added tests for the Khmer language (#4253)

This release alo bumps pathspec to v1 and fixes inconsistencies with Git's .gitignore logic (#4958). Now, files will be ignored if a pattern matches them, even if the parent directory is directly unignored. For example, Black would previously format exclude/not_this/foo.py with this .gitignore:

exclude/
!exclude/not_this/

Now, exclude/not_this/foo.py will remain ignored. To ensure exclude/not_this/ and all of it's children are included in formatting (and in Git), use this .gitignore:

*/exclude/*
!*/exclude/not_this/

This new behavior matches Git. The leading */ are only necessary if you wish to ignore

... (truncated)

• 6305bf1 Prepare 2026.1.0 release (#4892)
• e71305b Bump pypa/cibuildwheel from 3.3.0 to 3.3.1 (#4961)
• 21a2a8c Fix Shutdown multiprocessing Manager in schedule_formatting (#4952)
• e3146ce Bump docker/setup-buildx-action from 3.11.1 to 3.12.0 (#4919)
• fe1fbc4 Bump actions/upload-artifact from 5.0.0 to 6.0.0 (#4923)
• 2b4b7fc Bump actions/download-artifact from 6.0.0 to 7.0.0 (#4922)
• d745be6 docs: document --force-exclude for pre-commit workflows (#4957)
• b41acd6 Various CI and doc refactors (#4928)
• 6f43612 Handle pathspec v1 changes (#4958)
• 200c550 Bump furo from 2025.9.25 to 2025.12.19 in /docs (#4933)
• Additional commits viewable in compare view

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot merge will merge this PR after your CI passes on it
• @dependabot squash and merge will squash and merge this PR after your CI passes on it
• @dependabot cancel merge will cancel a previously requested merge and block automerging
• @dependabot reopen will reopen this PR if it is closed
• @dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore <dependency name> major version will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
• @dependabot ignore <dependency name> minor version will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
• @dependabot ignore <dependency name> will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
• @dependabot unignore <dependency name> will remove all of the ignore conditions of the specified dependency
• @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions