●Change8

v3.4.1

📦 sentence-transformers
✨ 1 features🐛 6 fixes🔧 5 symbols

Summary

Version 3.4.1 adds native Model2Vec support to SentenceTransformer and fixes several documentation and network‑request bugs.

Migration Steps

  1. Replace the previous StaticEmbedding loading pattern with direct SentenceTransformer initialization for Model2Vec models, e.g., `SentenceTransformer("model-id", device="cpu")`.
  2. No code changes are required for other models; the fix to `local_files_only` now correctly avoids network calls.

✨ New Features

  • Full Model2Vec integration: SentenceTransformer can now directly load Model2Vec models from the Hugging Face Hub without needing a StaticEmbedding module.

🐛 Bug Fixes

  • Fixed bug where using `local_files_only=True` still triggered a request to Hugging Face for model card metadata (PR #3202).
  • Fixed loss name documentation for CachedMultipleNegativesRankingLoss (PR #3191).
  • Updated jinja2 dependency from 3.1.4 to 3.1.5 in documentation builds (PR #3192).
  • Corrected typo in MegaBatchMarginLoss documentation (PR #3193).
  • Fixed type hint of `StaticEmbedding.__init__` (PR #3196).
  • Prevented call to `set_base_model` when `local_files_only=True` (PR #3202).

🔧 Affected Symbols

SentenceTransformerStaticEmbeddingCachedMultipleNegativesRankingLossMegaBatchMarginLossset_base_model