apparel23-qwen32b-kept-outfits-with-products
收藏资源简介:
该数据集名为“Apparel23 kept outfit bundles (with products)”,是Amazon Apparel 2023的outfit bundles(包含产品)集合。它作为DeepShopper项目中AMZ端的核心数据源,主要用于训练奖励模型V0(即qwen4b-apparel23-bundle-sft),同时为奖励模型V1提供正样本数据,生成Reducer的软黄金目标,并支持AMZ映射器的需求。数据集以JSONL格式提供,包含训练集(sft.train.jsonl)和测试集(sft.test.jsonl)。数据来源于公开的Amazon数据,仅限于研究用途。该数据集适用于文本生成、时尚推荐和outfit组合等任务,特别关注时尚领域的个性化推荐系统开发。
The dataset is named Apparel23 kept outfit bundles (with products) and is a collection of outfit bundles (including products) for Amazon Apparel 2023. It serves as the core data source for the AMZ side of the DeepShopper project, primarily used for training the reward model V0 (i.e., qwen4b-apparel23-bundle-sft), while also providing positive sample data for reward model V1, generating soft gold targets for the Reducer, and supporting the needs of the AMZ mapper. The dataset is provided in JSONL format, including a training set (sft.train.jsonl) and a test set (sft.test.jsonl). The data is sourced from publicly available Amazon data and is limited to research purposes. It is suitable for tasks such as text generation, fashion recommendation, and outfit combination, with a particular focus on the development of personalized recommendation systems in the fashion domain.
数据集概述
数据集名称:Apparel23 kept outfit bundles (with products)
许可证:other(其他)
语言:英语
任务类别:文本生成
标签:deepshopper, recommendation, fashion, outfit
来源:基于公开的 Amazon 数据
用途:研究用途
数据集详情
- 内容:包含 Amazon Apparel 2023 年的服装搭配包(outfit bundles),每个搭配包由一组带有角色标签的产品组成。
- 数据文件:
sft.{train,test}.jsonl,包含训练集和测试集。 - 相关项目:DeepShopper,该数据集用于其 AMZ 侧训练,具体包括:
- 训练奖励模型 V0(对应模型:
qwen4b-apparel23-bundle-sft)。 - 为奖励模型 V1 提供正样本种子。
- 为 Reducer 模型提供软黄金目标(soft-gold targets)。
- 提供 AMZ 映射(mapper)所需的数据。
- 训练奖励模型 V0(对应模型:
- 相关代码:https://github.com/clijo/reco-rl(分支:
outfit_bundle)。




