AI Blocks › Build and ship software
Embedding model trainer
Say what your search or matching must do. Get an embedding or reranker model trained from a ready template.
Embedding model trainer is a free AI skill for the job "Build and ship software". Give the instructions to an AI assistant that supports the required files and tools, then review its output.
Free · No account · CC BY 4.0 licence · Made by sickn33 · Level: expert · Thinking: high
Made by sickn33. See the original on GitHub; we copied it as it is, with its licence.
What it does
A trained search or matching model, built from a tested template rather than written from scratch. Say what your search must do and you get an embedding model or a reranker, fine-tuned on your data. The skill is a router. It asks whether you need a bi-encoder for fast retrieval, a cross-encoder to rerank the top results, or a sparse encoder for keyword-style search. Then it makes you read the matching guides on losses, evaluators and model design before touching code. You copy the production template script for that type and adapt it. It stresses matching the loss to your data shape and the right batch settings. You need Python, the sentence-transformers library and usually a GPU.
You give it
- What the model must find or match
- Training pairs or labelled data
- Base model choice (optional)
- A Python setup, ideally with a GPU
You get
- The right model type for your job
- A training script from the template
- Evaluation setup and metrics
- A fine-tuned model
Use it when
- Off-the-shelf search misses results in your field.
- You need a reranker to sort the top results.
- You have pairs of related texts to train on.
- You want to measure whether training helped.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.
- It works best when your AI can run code.