extract 0 0 0

erhwenkuo/hh_rlhf-chinese-zhtw

published by @erhwenkuo

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
1.2K
Last 30 days
64
Likes
10
Size
170.6 MB
Created on HF
2023-10-04
Age
3 years ago

Description

Dataset Card for "hh_rlhf-chinese-zhtw"

此數據集合併了下列的資料:

關於有用且無害的人類偏好數據,來自 Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback。這些數據旨在為後續 RLHF 訓練訓練偏好(或獎勵)模型。這些資料不用於對話代理人的監督訓練。根據這些資料訓練對話代理可能會導致有害的模型,這種情況應該避免。

人工生成並帶註釋的紅隊對話,來自減少危害的紅隊語言模型:方法、擴展行為和經驗教訓。這些數據旨在了解眾包紅隊如何建模以及哪些類型的紅隊攻擊成功或失敗。這些數據不用於微調或偏好建模(使用上面的數據進行偏好建模)。這些數據是從上述無害偏好建模數據導出的對話的完整轉錄本,其中僅將所選響應合併到整個轉錄本中。此外,文字記錄也透過人工和自動測量來標註整個對話的危害程度。

特別注意… See the full description on the dataset page: https://huggingface.co/datasets/erhwenkuo/hh_rlhf-chinese-zhtw.

Tags

task_categories:reinforcement-learninglanguage:zhlicense:mitsize_categories:100K<n<1Mformat:parquetmodality:textlibrary:datasetslibrary:pandaslibrary:mlcroissantlibrary:polarsarxiv:2204.05862region:us

README current version from Hugging Face


dataset_info:
features:

  • name: context
    list:
    • name: role
      dtype: string
    • name: text
      dtype: string
  • name: chosen
    struct:
    • name: role
      dtype: string
    • name: text
      dtype: string
  • name: rejected
    struct:
    • name: role
      dtype: string
    • name: text
      dtype: string
      splits:
  • name: train
    num_bytes: 302431699
    num_examples: 344317
    download_size: 178897699
    dataset_size: 302431699
    configs:
  • config_name: default
    data_files:
    • split: train
      path: data/train-*
      license: mit
      task_categories:
  • reinforcement-learning
    language:
  • zh

Dataset Card for "hh_rlhf-chinese-zhtw"

此數據集合併了下列的資料:

  1. 關於有用且無害的人類偏好數據,來自 Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback。這些數據旨在為後續 RLHF 訓練訓練偏好(或獎勵)模型。這些資料不用於對話代理人的監督訓練。根據這些資料訓練對話代理可能會導致有害的模型,這種情況應該避免。
  2. 人工生成並帶註釋的紅隊對話,來自減少危害的紅隊語言模型:方法、擴展行為和經驗教訓。這些數據旨在了解眾包紅隊如何建模以及哪些類型的紅隊攻擊成功或失敗。這些數據不用於微調或偏好建模(使用上面的數據進行偏好建模)。這些數據是從上述無害偏好建模數據導出的對話的完整轉錄本,其中僅將所選響應合併到整個轉錄本中。此外,文字記錄也透過人工和自動測量來標註整個對話的危害程度。

特別注意

數據(尤其是無害偏好數據和紅隊數據)包含可能令人反感或令人不安的內容。主題包括但不限於歧視性語言以及對虐待、暴力、自殘、剝削和其他可能令人不安的主題的討論。

請僅根據您個人的風險承受能力來處理資料。這些數據旨在用於研究目的,特別是可以使模型「減少」危害的研究。

如上所述,這些資料「不」該用於訓練對話代理,因為這可能會導致有害的模型行為。

原始數據

原作者在 GitHub 上託管此資料集:https://github.com/anthropics/hh-rlhf

轉譯來源

本數據集來自 dikw/hh_rlhf_cn 并使用了 OpenCC 來進行簡繁轉譚。

基於 Anthropic 論文 Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback 開源的 helpful 和 harmless 數據,使用翻譯工具進行了翻譯。

  • hh_rlhf_train.jsonl 合併中英文訓練集資料 清洗過後 17 萬條
  • hh_rlhf_test.jsonl 合併中英文測試集資料 清洗過後 9 千條
  • harmless_base_cn_train.jsonl 42394 條
  • harmless_base_cn_test.jsonl 2304 條
  • helpful_base_cn_train.jsonl 43722 條
  • helpful_base_cn_test.jsonl 2346 條

README history 4 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2023-10-04Update README.md643b4ae2.9 KB
    Loading...
  2. 2023-10-04Upload README.md with huggingface_hub4367e90733 B
    Loading...
  3. 2023-10-04Upload data/train-00000-of-00001-42ca637f511dffc4.parquet with huggingface_huba3dd93a0 B
    Loading...
  4. 2023-10-04initial commitcd1def10 B
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Related datasets same stage · extract

Read further

Metadata

License
mit

Repository

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python)
from datasets import load_dataset
ds = load_dataset("erhwenkuo/hh_rlhf-chinese-zhtw")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="erhwenkuo/hh_rlhf-chinese-zhtw", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/erhwenkuo/hh_rlhf-chinese-zhtw
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration