| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
[`feat`] Add model merging via mergekit for all archetypes (#3776) * Add model merging via mergekit * Bump mergekit minimum version to 0.1.4 Co-authored-by: Yongbin Choi <whybe.choi@gmail.com> * Simplify model merging with mergekit's in-memory API * Run merge tests across CI and verify small-tensor pruning * Add model merging guides for all model types * Reject incompatible model configurations during merging --------- Co-authored-by: Yongbin Choi <whybe.choi@gmail.com> Co-authored-by: Tom Aarsen <37621491+tomaarsen@users.noreply.github.com> Co-authored-by: Tom Aarsen <Cubiegamedev@gmail.com> | 9 天前 | |
Document Hugging Face login and refresh retrieval examples (#4160) | 18 天前 | |
[`docs`] Use modality-neutral terms (input, document) in loss docs & docstrings (#3772) * [`docs`] Use modality-neutral terms (input, document) in loss docs & docstrings * Fix table markdown formatting * Apply suggestions, remove InputExample in docstrings | 5 个月前 | |
Add SlideOn and mSlideOn to Sparse Encoder pretrained models (#4234) | 10 天前 | |
Correct the DDP evaluator note in the training overviews (#4248) * Correct the DDP evaluator note in the training overviews Since #4004 the trainer shares evaluator inference across DDP processes, computes and writes metrics on process zero, and broadcasts them, as the distributed training guide describes. The four training overviews still said the evaluator only runs on the first device. Replace that warning with a note matching the guide, and point the CrossEncoder link at the existing sentence_transformer distributed page. * Fold DDP evaluator guidance into the overview text --------- Co-authored-by: Tom Aarsen <Cubiegamedev@gmail.com> | 6 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 9 天前 | ||
| 18 天前 | ||
| 5 个月前 | ||
| 10 天前 | ||
| 6 天前 |