pricer-data-small / README.md
saxon11's picture
Update README.md
39eaa6e verified
metadata
dataset_info:
  features:
    - name: text
      dtype: string
    - name: price
      dtype: float64
  splits:
    - name: train
      num_bytes: 16062338
      num_examples: 20000
    - name: test
      num_bytes: 6829484
      num_examples: 8544
  download_size: 11195441
  dataset_size: 22891822
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
license: mit
task_categories:
  - text-generation
language:
  - en
size_categories:
  - 10K<n<100K
tags:
  - price-prediction
  - product-descriptions
  - fine-tuning
  - e-commerce
pretty_name: Pricer Data (Small)

Pricer Data (Small)

A dataset of product descriptions paired with their prices, designed for fine-tuning language models to predict product prices from text descriptions.

Dataset Summary

Each example contains a natural language prompt asking "How much does this cost to the nearest dollar?" followed by a product title, description, and attributes. The target is the actual price as a float value.

Dataset Structure

Split Rows
Train 20,000
Test 8,544
Total 28,544

Fields

  • text (string): A prompt containing the product name, description, specifications, and the prefix "Price is $" for the model to complete.
  • price (float64): The actual price of the product in USD.

Example

{
  "text": "How much does this cost to the nearest dollar?\n\n[Product title and description]\n\nPrice is $",
  "price": 82.09
}

Data Distribution

  • Text lengths: Primarily between 483–580 characters (2.2% variance)
  • Price range: Most prices fall between $1.11 and $101.00 (87.3% of data)

Usage

from datasets import load_dataset

dataset = load_dataset("saxon11/pricer-data-small")
train = dataset["train"]
test = dataset["test"]

Intended Use

This dataset was created to fine-tune a Llama 3.1 8B model using QLoRA and SFTTrainer for the task of product price prediction. It can be used for any text-to-number regression task framed as language model completion.

Source

Product descriptions and prices sourced from e-commerce listings, primarily covering appliance parts, accessories, and home goods.

Author

Josh Janzen · joshjanzen.com