no-robots-sharegpt / README.md
Doctor-Shotgun's picture
Update the configuration to use 'load_dataset' properly. (#3)
ff40b93
|
raw
history blame
852 Bytes
metadata
license: cc-by-nc-4.0
configs:
  - config_name: default
    data_files: no-robots-sharegpt-fixed.jsonl
task_categories:
  - text-generation
  - conversational
language:
  - en

no-robots-sharegpt

HuggingFaceH4/no_robots with both test and train splits combined and converted to ShareGPT format for use in common training repositories.

Please refer to the original repository's dataset card for more information.

no-robots-sharegpt.jsonl

  • Original dataset converted to ShareGPT

no-robots-sharegpt-fixed.jsonl

  • Manual edits were made to ~10 dataset entries that were throwing warnings in axolotl - turns out that some of the multi-turn conversations had incorrectly-parsed turns in the original dataset

Auto-converted Parquet is utilizing the no-robots-sharegpt-fixed.jsonl file.