datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
dolly_hhrlhf-text2text
dolly_hhrlhf-text2text
This is mosaicml/dolly_hhrlhf with the following changes:
clean up/adapt prompt column for the text2text-generation task (no need for a special template)
split the original train set into a 95% train and an explicit validation set (5%)
fixed extra spaces in puncuation (as this is not a French dataset)
details on extra spaces:
Original sentence 1: How can I be healthy ?
Fixed sentence 1: How can I be healthy?
srl_datasets_text2text_samplegoodwiki-text2text-completion
goodwiki-text2text-completion
Wikipedia articles split on paragraphs and/or markdown headers pseudo-randomly.
inputs/outputs are in columns inputs/targets
see the instruct config with prompts prepended to the text in inputs
Source data: euirim/goodwiki
dataset_info:
features:
- name: pageid
dtype: int64
- name: title
dtype: string
- name: revid
dtype: int64
- name: description
dtype: string
- name: categories
sequence: string
- name: inputs… See the full description on the dataset page: https://huggingface.co/datasets/pszemraj/goodwiki-text2text-completion.text2text-generation-checkpoint-downloadsStylumia-nxt-text2texttext2text
Dataset mimba/text2text
📝 Description
This dataset provides multilingual parallel sentence pairs for machine translation (text-to-text tasks).Currently, it includes Ngiemboon ↔ French (40,968 examples).In the future, additional language pairs will be added (e.g., Ngiemboon ↔ English, etc.).
Total examples (current): 40,968
Columns:
source_text: source sentence
target_text: target sentence
source_lang: ISO 639‑3 language code of the source (e.g., nnh)… See the full description on the dataset page: https://huggingface.co/datasets/mimba/text2text.bgibr-text2text