skills.rest
MS33834
MS33834COMMUNITY
Explore all skills by this author →
GitHub

Embedding Model Training

Select encoder type, loss function, and evaluator for embedding model training.

Updated Aug 23, 2026
Search Tags:#fine-tuning#retrieval#embedding#sentence-transformers#reranker#bi-encoder#splade
One-click install
npx skills add https://github.com/MS33834/AI-SKILL --skill embedding-model-training
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Embedding Model Training
Source: https://github.com/MS33834/AI-SKILL/tree/main/skills/embedding-model-training
Command: npx skills add https://github.com/MS33834/AI-SKILL --skill embedding-model-training

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you choose and configure the right training approach when you need to fine-tune an embedding model instead of using a pretrained one unchanged. It reduces guesswork around model family, loss function, evaluation setup, and training safeguards.

Core Features & Use Cases

  • Model selection: Picks the right encoder type for retrieval, similarity, reranking, classification, or sparse search.
  • Training guidance: Maps your data shape, such as pairs or triples, to an appropriate loss and evaluator.
  • Practical guardrails: Highlights common training footguns like duplicate negatives, incorrect activation settings, and unsafe checkpointing choices.
  • Use case: A team wants to adapt a retrieval model for legal documents and needs a reliable plan for base model choice, hard negatives, and evaluation metrics.

Quick Start

Ask for a training plan for your embedding task, including the encoder type, data shape, base model, loss, evaluator, and key training pitfalls.

Frequently Asked Questions about Embedding Model Training

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I choose the right loss function for fine-tuning an embedding model?▼

Picking the right encoder for fine-tuning an embedding model depends on your specific workflow: bi-encoders for retrieval, cross-encoders for reranking, and sparse-encoders for learned sparse retrieval. Select the base model that matches your target task.

How do I configure evaluators and training guardrails for sentence-transformers?▼

Configuring evaluators and training guardrails for sentence-transformers requires selecting appropriate metrics for your data shape and avoiding common footguns like duplicate negatives, incorrect activation settings, and unsafe checkpointing choices.

What's the best way to train a SPLADE model for learned sparse retrieval?▼

Training a SPLADE model for learned sparse retrieval involves selecting a sparse-encoder setup, mapping your data shape to an appropriate loss function, and applying training guardrails to prevent incorrect activation settings during fine-tuning.

Why does my retrieval model performance drop after fine-tuning?▼

Retrieval model performance may drop after fine-tuning due to common training footguns like duplicate negatives or unsafe checkpointing choices. Selecting the correct evaluator and activation settings prevents these degradation issues.

Related Skills

Explore More Software Engineering Skills→
agentic-inagentic-in
577

Training

Plan fine-tuning or post-training workflows from objectives, data, and hardware constraints.

Official
Advanced
Devsoul2026Devsoul2026
150

clip

Classify images by text labels using a pre-trained vision-language model.

Community
Advanced
nakano1122nakano1122
2

search-model-builder

Guide construction and fine-tuning of HuggingFace sentence-transformers search models.

Community
Advanced
xiaoquqixiaoquqi

clip

Embed images and text with CLIP for zero-shot classification and retrieval.

Community
Intermediate