Datasets:

fermsi
/

berkeley-na-brewing-corpus

license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - brewing
  - non-alcoholic-beer
  - berkeley-yeast
  - nanochat
size_categories:
  - 1K<n<10K

Berkeley NA Brewing Corpus

A text corpus about non-alcoholic beer brewing, created for training with Karpathy's nanochat.

Dataset Description

This dataset contains technical information about brewing non-alcoholic beer using Berkeley Yeast's maltose-negative yeast strains. The content covers:

Food safety best practices for NA beer production
Recipe development and mash procedures
Yeast selection and fermentation techniques
Pasteurization and packaging guidelines
Hop usage and flavor optimization

Sources

The corpus is compiled from:

Berkeley Yeast technical documentation
Video transcripts from brewing tutorials
Webinar content on NA beer production

Usage

For nanochat pretraining

# Load the raw text
with open("train.txt", "r") as f:
    corpus = f.read()

With Hugging Face datasets

from datasets import load_dataset

dataset = load_dataset("YOUR_USERNAME/berkeley-na-brewing-corpus")

Format

train.txt - Combined plain text corpus
train.jsonl - JSON Lines format with "text" field
train.parquet - Parquet format (if available)

License

MIT License