README.md

metadata

language:
  - en
license: apache-2.0
multilinguality:
  - monolingual
source_datasets:
  - bartman081523/stable-diffusion-discord-prompts
  - succinctly/midjourney-prompts
  - Gustavosta/Stable-Diffusion-Prompts
pretty_name: multi text2image prompts a dataset collection
tags:
  - text generation
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
  - config_name: original
    data_files:
      - split: train
        path: original/train-*
      - split: test
        path: original/test-*
dataset_info:
  - config_name: default
    features:
      - name: text
        dtype: string
      - name: src_dataset
        dtype: string
    splits:
      - name: train
        num_bytes: 262736830
        num_examples: 1677221
      - name: test
        num_bytes: 56294291
        num_examples: 292876
    download_size: 151054782
    dataset_size: 319031121
  - config_name: original
    features:
      - name: text
        dtype: string
      - name: src_dataset
        dtype: string
    splits:
      - name: train
        num_bytes: 741427383
        num_examples: 3551734
      - name: test
        num_bytes: 83615440
        num_examples: 399393
    download_size: 402186258
    dataset_size: 825042823
task_categories:
  - text-generation
  - feature-extraction

text2image multi-prompt(s): a dataset collection

• collection of several text2image prompt datasets
• data was cleaned/normalized with the goal of removing "model specific APIs" like the "--ar" for Midjourney and so on
• data de-duplicated on a basic level: exactly duplicate prompts were dropped (after cleaning and normalization)

updates

• Oct 2023: the default config has been updated with better deduplication. It was deduplicated with minhash (params: n-gram size set to 3, deduplication threshold at 0.6, hash function chosen as xxh3 with 32-bit hash bits, and 128 permutations with a batch size of 10,000.) which drops 2+ million rows.
  • original version is still available under config_name="original"

contents

default:

DatasetDict({
    train: Dataset({
        features: ['text', 'src_dataset'],
        num_rows: 1677221
    })
    test: Dataset({
        features: ['text', 'src_dataset'],
        num_rows: 292876
    })
})

For original config:

DatasetDict({
    train: Dataset({
        features: ['text', 'src_dataset'],
        num_rows: 3551734
    })
    test: Dataset({
        features: ['text', 'src_dataset'],
        num_rows: 399393
    })
})

NOTE: as the other two datasets did not have a validation split, the validation split of succinctly/midjourney-prompts was merged into train.