| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
Add training skill: `train-sentence-transformers` (#3752) * Add 3 training skills * Merge into one skill & try 'less is more' strategy * Harden skill templates: smoke-test, metric keys, Normalize fix * Log actual pushed URL * Add VERDICT in variant templates + SPARSE keys + VLM precision * workflow check, listwise eval, multilingual scoring, SKILL.md cross-refs * Tighten SKILL.md description * Rename to plural train-sentence-transformers * Add mentions of 'hf skills add train-sentence-transformers' | 4 个月前 | |
docs: Add docs analytics (#3577) | 9 个月前 | |
docs: fix Constrastive/heaqp typos and subject-verb agreement (#3983) | 3 天前 | |
[docs] Name the metric on the quality axis of the backend benchmark figures (#3977) The bottom subfigure was labelled "Performance Ratio", which reads as a second speed measure rather than the retrieval quality it plots. It now names the metrics that actually fed it: NanoBEIR NDCG@10 for the Cross Encoder, Sparse Encoder and Multi-Vector Encoder figures, and NanoBEIR NDCG@10 plus STSb Spearman for the Sentence Transformer ones. The figure title and the surrounding prose say "quality" instead of "performance" to match. All eight figures are re-rendered from unchanged benchmark results, so every number is identical to before. Closes #3950 | 7 天前 | |
[docs] Name the metric on the quality axis of the backend benchmark figures (#3977) The bottom subfigure was labelled "Performance Ratio", which reads as a second speed measure rather than the retrieval quality it plots. It now names the metrics that actually fed it: NanoBEIR NDCG@10 for the Cross Encoder, Sparse Encoder and Multi-Vector Encoder figures, and NanoBEIR NDCG@10 plus STSb Spearman for the Sentence Transformer ones. The figure title and the surrounding prose say "quality" instead of "performance" to match. All eight figures are re-rendered from unchanged benchmark results, so every number is identical to before. Closes #3950 | 7 天前 | |
docs: Point cross-references at their documented targets (#3947) * docs: Point cross-references at their documented targets * docs: Resolve two more cross-reference clusters | 21 天前 | |
[`docs`] Use direct class imports in examples & docs (drop `losses.MSELoss(...)` style) (#3770) * [`docs`] Use direct class imports in examples & docs (drop `losses.MSELoss(...)` style) * Fix 2 unrelated typos Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> | 3 个月前 | |
[docs] Name the metric on the quality axis of the backend benchmark figures (#3977) The bottom subfigure was labelled "Performance Ratio", which reads as a second speed measure rather than the retrieval quality it plots. It now names the metrics that actually fed it: NanoBEIR NDCG@10 for the Cross Encoder, Sparse Encoder and Multi-Vector Encoder figures, and NanoBEIR NDCG@10 plus STSb Spearman for the Sentence Transformer ones. The figure title and the surrounding prose say "quality" instead of "performance" to match. All eight figures are re-rendered from unchanged benchmark results, so every number is identical to before. Closes #3950 | 7 天前 | |
[docs] Name the metric on the quality axis of the backend benchmark figures (#3977) The bottom subfigure was labelled "Performance Ratio", which reads as a second speed measure rather than the retrieval quality it plots. It now names the metrics that actually fed it: NanoBEIR NDCG@10 for the Cross Encoder, Sparse Encoder and Multi-Vector Encoder figures, and NanoBEIR NDCG@10 plus STSb Spearman for the Sentence Transformer ones. The figure title and the surrounding prose say "quality" instead of "performance" to match. All eight figures are re-rendered from unchanged benchmark results, so every number is identical to before. Closes #3950 | 7 天前 | |
Rename model doc pages to consistent model.md and add redirects | 5 个月前 | |
[`docs`] Rewrite the https://sbert.net documentation for v3.0 (#2632) * Start restructuring/rewriting the docs * Update Pretrained Models section for ST * Update & add many docstrings * Completely overhaul "Training Overview" docs page for ST * Update dataset overview * Remove kwargs from paraphrase_mining signature * Add "aka sbert" * Remove Hugging Face docs page * Update ST Usages * Fix some links * Use the training examples corresponding to that model type * Add hyperparameter optimization example script + docs * Add distributed training docs * Complete rewrite for the Sentence Transformer docs portion * Update the CE part of the docs * Specify if __name__ == "__main__" & dataloader_drop_last with DDP * Update the entire project to Google-style docstring * Remove contact page * Update README with updated links, etc. * Update the loss examples * Fix formatting * Add remove_columns/select_columns tip to dataset overview | 2 年前 | |
[`docs`] Fix doc build problems (part 1) (#3811) * Docs: Add missing `sphinx_rtd_theme` dependency to `docs` group This resolves a local docs build error: ```console $ python -m venv venv-docs $ source venv-docs/bin/activate $ python -m pip install --group=docs $ make docs ... no theme named 'sphinx_rtd_theme' found (missing theme.toml?) ... $ deactivate ``` * Docs: Update a redirecting PyTorch intersphinx mapping This resolves an info message when building the docs: ``` intersphinx inventory has moved: https://pytorch.org/docs/stable/objects.inv -> https://docs.pytorch.org/docs/stable/objects.inv ``` * Docs: Resolve two autodoc warnings The warnings are: ``` WARNING: missing attribute modality_config in object sentence_transformers.base.modules.Transformer [autodoc] WARNING: missing attribute transformer_task in object sentence_transformers.base.modules.Transformer [autodoc] ``` While these attributes exist on instances, they are not methods nor properties. * Docs: Resolve a docutils warning The warning is: ``` sentence_transformers/sentence_transformer/losses/multiple_negatives_ranking.py: docstring of sentence_transformers.sentence_transformer.losses.multiple_negatives_ranking.MultipleNegativesRankingLoss:52: ERROR: Unexpected indentation. [docutils] ``` RST requires nested lists to be separated by blank lines. * Docs: Resolve a docutils warning The warning is: ``` sentence_transformers/sentence_transformer/losses/cached_multiple_negatives_ranking.py: docstring of sentence_transformers.sentence_transformer.losses.cached_multiple_negatives_ranking.CachedMultipleNegativesRankingLoss:46: ERROR: Unexpected indentation. [docutils] ``` RST requires nested lists to be separated by blank lines. * Docs: Resolve a Sphinx TOC warning The warning is: ``` docs/package_reference/sparse_encoder/index.rst:4: WARNING: toctree contains reference to nonexisting document 'docs/package_reference/sentence_transformer/sampler' [toc.not_readable] ``` The `sampler.md` file was moved in commit `4cdbd836717e420e3b4c2ea72c894fb1e2120100`. * Docs: Resolve a docutils warning The warning is: ``` sentence_transformers/sparse_encoder/model.py: docstring of sentence_transformers.sparse_encoder.model.SparseEncoder.decode:12: ERROR: Unexpected indentation. [docutils] ``` The root issue was the formatting of the conditional return types. Sphinx could not handle two type annotations on separate lines, and ultimately rendered the content incorrectly. * Docs: Resolve 3 docutils warnings The warnings are: ``` index.rst:7: WARNING: Duplicate explicit target name: "quickstart". [docutils] index.rst:7: WARNING: Duplicate explicit target name: "quickstart". [docutils] index.rst:17: WARNING: Duplicate explicit target name: "quickstart". [docutils] ``` The issue is that links created with a single trailing underscore are registered with Sphinx. The solution is to use two trailing underscores which create anonymous links. This avoids the "duplicate explicit target name" warning. * Anonymize 4th quickstart + update sphinx_rtd_theme version Co-authored-by: Tom Aarsen <37621491+tomaarsen@users.noreply.github.com> --------- Co-authored-by: Tom Aarsen <37621491+tomaarsen@users.noreply.github.com> | 2 个月前 | |
Add a v6 breaking-changes section to the migration guide, drop the dead transformers 4.41 gates Also generate widget examples for set_transform datasets | 29 天前 | |
tests: Fix the pretrained MultiVectorEncoder tests for device tensors and Hub changes (#3949) * tests: Fix the pretrained MultiVectorEncoder tests for device tensors and Hub changes * Keep the -hf example in the ForRetrieval paragraph, mark LFM2 as self-pinned The colpali-engine eval-rst block describes the transformers-native *ForRetrieval auto-detection path, so its example has to stay on vidore/colqwen2-v1.0-hf. vidore/colqwen2-v1.0 now ships a Sentence Transformers configuration, so its projection runs in a Dense module rather than inside the model. The migration table below it still points at vidore/colqwen2-v1.0. MODELS_TO_MAXSIM no longer claims PyLate for the LFM2 row: PyLate sums MaxSim in the checkpoint's bfloat16 while v6 upcasts to float32, so that row pins our own output rather than a cross-library reference. | 21 天前 | |
[`chore`] Rename master to main, update outdated URLs (#3579) * Rename master to main, update outdated URLs * Update some cases of UKPLab not yet being replaced by huggingface | 9 个月前 | |
docs: Point cross-references at their documented targets (#3947) * docs: Point cross-references at their documented targets * docs: Resolve two more cross-reference clusters | 21 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 4 个月前 | ||
| 9 个月前 | ||
| 3 天前 | ||
| 7 天前 | ||
| 7 天前 | ||
| 21 天前 | ||
| 3 个月前 | ||
| 7 天前 | ||
| 7 天前 | ||
| 5 个月前 | ||
| 2 年前 | ||
| 2 个月前 | ||
| 29 天前 | ||
| 21 天前 | ||
| 9 个月前 | ||
| 21 天前 |