#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/dataset-preparation-curation/the-alpaca-format

░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░███████████████████████████████████████████████████████████████████████████████████████████

#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/dataset-preparation-curation/the-alpaca-format

Repovive

The Alpaca Format - Dataset Preparation & Curation | LLM Fine-Tuning | Repovive

Repovive.

LLM Fine-Tuning10 sections · 313 units1h total

Open in Course

The Alpaca Format

Alpaca format includes $3$ fields:

instruction: What the model should do
input: Additional context (optional)
output: The expected response

{
  "instruction": "Classify the sentiment",
  "input": "This product exceeded my expectations!",
  "output": "Positive"
}

Many tools and datasets use this format from Stanford's Alpaca project. If you're unsure which format to use, Alpaca is a safe default that most frameworks support.