HYUKI5's picture
nvidia/Nemotron-RL-coding-competitive_coding@755d5910fc86 を複製し、カードを日本語で書く
f6ec219 verified
|
Raw History Blame Contribute Delete
9.73 kB
metadata
license: cc-by-sa-4.0
pretty_name: nvidia/Nemotron-RL-coding-competitive_coding (mirror)
source_datasets:
  - nvidia/Nemotron-RL-coding-competitive_coding
tags:
  - laya
  - laya-finetune
  - mirror
  - domain-code
  - use-conditional
  - genre-exam
  - label-text
  - annot-natural
  - lang-en
language:
  - en
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: responses_create_params
      struct:
        - name: input
          list:
            - name: role
              dtype: string
            - name: content
              dtype: string
    - name: verifier_metadata
      struct:
        - name: unit_tests
          struct:
            - name: inputs
              list: string
            - name: outputs
              list: string
    - name: hash_id
      dtype: string
    - name: dataset
      dtype: string
    - name: source
      dtype: string
  splits:
    - name: train
      num_bytes: 5280869002
      num_examples: 16083
  download_size: 2933275582
  dataset_size: 5280869002

nvidia/Nemotron-RL-coding-competitive_coding(複製)

データの内容

Python の競技プログラミングの問題と、そのユニットテストを集めた、コーディングの RL 用のデータ。1 行が 1 件の問題で、問題の依頼は responses_create_params、採点用のテストは verifier_metadata の unit_tests に入る。問題とテストケースは CodeContests と Open-R1(codeforces)から集めたもの。

データの概要

項目 値
行数 16,083(1 構成・1 分割の合計)
列数 5
容量 15.68 GB(ファイル 15 本)
ファイルの形式 parquet 11、jsonl 2
言語 en
公開 public

構成と分割

構成(config) 分割(split)ごとの行数 列数
default train 16,083 5

列(構成 default・分割 train)

列 型 先頭の行の値
responses_create_params struct{input: list<struct{…}>} {input}
verifier_metadata struct{unit_tests: struct{inputs: list<string>, outputs: list<string>}} {unit_tests}
hash_id string fd52e18df463da29e06dab4545f6477b
dataset string open-r1/codeforces
source string codeforces

行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は HYUKI5/dataset-catalog にある。

複製について

nvidia/Nemotron-RL-coding-competitive_coding をリビジョン 755d5910fc8646b385e3926eec08c152051cdc07 の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。

出典

論文

  • 元のカードに論文へのリンクは見つからなかった

アノテーションの方法

データセットの作成日:

2025 年 7 月 11 日

微調整に使うときの注意

  • 調査メモの台帳での区分: B(継承条件・gated・モデル由来の条件・上流の規約などが付く)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
  • Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
  • 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。

調査メモの記述

メモ 節 ライセンス(記載) 確認 注記
nvidia_nemotron.md 付録. 取得した nvidia 組織のデータセット全件の簡易一覧 (L296) cc-by-sa-4.0 確認済 確認印なし(各ファイルの凡例により確認済扱い) / 原文注記: × コード・パズル・化学

元のカード(日本語訳)

以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。


データセットの説明:

Nemotron-RL-coding-competitive_coding データセットは、Python のみを対象とした、推論に基づく合成データセットである。競技プログラミング形式の問題と、そのユニットテストを含む。これらの問題とテストケースは、CodeContests(deepmind/code_contests)と Open-R1(open-r1/codeforces)から収集された。

このデータセットは、NVIDIA の NeMo Gym の一部として公開されている。NeMo Gym は、大規模言語モデルを学習させるための強化学習環境を構築するフレームワークである。NeMo Gym には、検証可能な報酬による強化学習(RLVR)を可能にする学習環境とデータセットのコレクションが増え続けている。

NeMo Gym は、NVIDIA NeMo フレームワーク に含まれるオープンソースのライブラリである。NVIDIA NeMo フレームワークは、GPU で高速化されたエンドツーエンドの学習フレームワークであり、大規模言語モデル(LLM)、マルチモーダルモデル、音声モデルを対象とする。

このデータセットは Hugging Face 組織 の一部である。

このデータセットは商用利用できる。

データセットの所有者:

NVIDIA Corporation

データセットの作成日:

2025 年 7 月 11 日

ライセンス/利用規約:

このデータセットは、Creative Commons Attribution 4.0 International License(CC-BY-4.0)の下で提供される。

想定される用途:

NeMo Gym と併せて、LLM の事後学習に使う。

データセットの特徴

** データ収集方法

  • [合成]

** ラベリング方法

  • [自動]

データ形式

テキストのみ。NeMo Gym と互換性がある。

データ量

レコード数: 25k 件のサンプル データ容量: 7.62 GB

参考文献:

NeMo Gym

倫理的な配慮:

NVIDIA は、信頼できる AI は共有の責任であると考えており、幅広い AI 用途の開発を可能にするため、方針と実践を確立している。利用規約に従ってダウンロードまたは利用される場合、開発者は社内のモデルチームと協力して、このモデルが該当する業界と用途の要件を満たし、想定外の製品の誤用に対処することを確認されたい。 モデルの品質、リスク、セキュリティ上の脆弱性、または NVIDIA の AI に関する懸念は、こちらから報告されたい。

原文(英語)

Dataset Description:

The Nemotron-RL-coding-competitive_coding dataset is a python-only, reasoning-based, synthetic dataset. It contains competitive coding style problems and their unit test cases. These questions and test cases are collected from CodeContests (deepmind/code_contests), and Open-R1 (open-r1/codeforces) .

This dataset is released as part of NVIDIA NeMo Gym, a framework for building reinforcement learning environments to train large language models. NeMo Gym contains a growing collection of training environments and datasets to enable Reinforcement Learning from Verifiable Reward (RLVR).

NeMo Gym is an open-source library within the NVIDIA NeMo framework, NVIDIA's GPU accelerated, end-to-end training framework for large language models (LLMs), multi-modal models and speech models.

This dataset is part of the Hugging Face Org.

This dataset is ready for commercial use.

Dataset Owner(s):

NVIDIA Corporation

Dataset Creation Date:

11 Jul 2025

License/Terms of Use:

This dataset is governed by the Creative Commons Attribution 4.0 International License (CC-BY-4.0)

Intended Usage:

To be used with NeMo Gym for post-training LLMs.

Dataset Characterization

** Data Collection Method

  • [Synthetic]

** Labeling Method

  • [Automated]

Dataset Format

Text Only, Compatible with NeMo Gym

Dataset Quantification

Record Count: 25k samples Data Storage: 7.62 GB

Reference(s):

NeMo Gym

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.