W-61/llama-multi-beta-testing-margin_dpo-hh-helpful-beta-0p5-20260429-085449-margin
收藏Hugging Face2026-04-29 更新2026-05-03 收录
下载链接：
https://hf-mirror.com/datasets/W-61/llama-multi-beta-testing-margin_dpo-hh-helpful-beta-0p5-20260429-085449-margin
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是从Margin-DPO训练运行中导出的每步边缘摘要统计信息。包含了训练过程中的多个统计特征，如epoch、step、batch_size、mean、std等，以及每个步骤的边缘值样本。数据集来源于特定的训练运行，使用了Anthropic/hh-rlhf数据混合器，并记录了训练参数如beta值为0.5，f_divergence_type为reverse_kl等。
Per-step margin summary statistics exported from a Margin-DPO training run. Includes various statistical features during training such as epoch, step, batch_size, mean, std, etc., as well as per-example margins for each logged step. The dataset originates from a specific training run using the Anthropic/hh-rlhf dataset mixer, with recorded training parameters like beta=0.5, f_divergence_type=reverse_kl, etc.
提供机构：
W-61
