README.md

metadata

license: mit
dataset_info:
  features:
    - name: text
      dtype: string
    - name: chat
      dtype: string
  splits:
    - name: train
      num_bytes: 48824063
      num_examples: 83842
  download_size: 23302075
  dataset_size: 48824063
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Promptvieh dataset

83k aesthetically pleasing StableDiffusion prompts

text = text completion template

chat = chat completion template

User input "prompt!" generates a general prompt, and "prompt anime!" a danbooru tag based prompt