Sentence Transformers Training

Train retrieval embeddings and rerankers on your own data.

The Agent selects a model, loss and evaluation method, prepares negative examples and starts from upstream training templates. Supports embedding models and CrossEncoder rerankers when general-purpose retrieval performs poorly on your products, documents or questions.

Example task

Fine-tune a reranker using these query-document relevance labels. Keep an independent test set and compare NDCG@10 before and after training.

Upstream project ↗

View or edit installation request

Send the request to your Agent. It includes the original source, installation scope and a verification task.

View or edit installation request

Please install and help me use Sentence Transformers Training (huggingface) in my current agent environment.
Upstream: https://github.com/huggingface/skills
Exact source: https://github.com/huggingface/skills/tree/f3186efbbc322121eb5d0f31e8a1d669ee961159/skills/train-sentence-transformers
Installation guide: https://github.com/huggingface/skills#readme
Reviewed source revision: f3186efbbc322121eb5d0f31e8a1d669ee961159 (2026-09-16)
Installation unit: Complete skills/train-sentence-transformers with references and scripts/train_*_example.py. Install matching sentence-transformers and training dependencies from the template.
Minimal verification: Run training and evaluation on a small sample, keep training and test data separate, and save baseline and final metrics.

Read the upstream instructions, check compatibility and required resources, and prefer project scope where supported. Install the complete skill directory or required tool package, including mandatory dependencies. If the current revision differs, check the changes first. Keep existing configuration and explain any conflicting setup. Ask me for credentials only if needed, in my local agent session. Verify discovery/connection and run a minimal task; report what was actually verified and how to use it.

My task:

Sources & verification2026-09-16

Publisher casesAuthor used this Skill to start Ettin reranker training ↗

The author explicitly credits this Skill for the initial training script and publishes the model, data, recipe and evaluation. The model result is not a general measure of Skill effectiveness.

Installation guide ↗Reviewed source ↗

Source revision: f3186efbbc32
License: Apache-2.0
Repository stars: 11060

My reading

Start a learning journey

Your next good read

Pick up where you left off

Sentence Transformers Training

Example task