{% extends "_base.html" %} {% block content %}
Workflow · dataset

Build a dataset

Thirteen templates to turn pairs, docs, code, seeds, and HF datasets into JSONL the fine-tune workflow can consume — including DPO preference pairs. The LLM-driven templates call this Lab's API.

{# Step 1: pick template #}

Pick a template

{# Step 2: fill inputs #}

{# Step 3: run #}

Generating


  
{# Step 4: save / push #}

Done

Dataset written to:


    

Point Fine-tune → Dataset at to use it.

Push to Hugging Face as a dataset

Build another
{% endblock %}