相关数据集
cygu/sampling-distill-train-data-kth-shift4
该数据集用于基于采样的水印蒸馏训练,使用了KTH(s=4)水印策略。数据集包含640,000个水印样本,每个样本长度为256个token,这些样本是基于Llama 2 7B模型生成的。每个样本使用来自OpenWebText的50-token前缀进行提示,但这些前缀不包含在样本中。
Hugging Face2024-05-23 更新220
PolarisEvals/training_criteria_dpo_distill_completeness_2stage_questions
--- dataset_info: features: - name: query dtype: string - name: chunk dtype: string - name: justification dtype: string - name: question dtype: string splits: - name: tra
Hugging Face2024-06-05 更新150
Devparker-2002/flanv2_deduplicated_filtered_dataset
--- dataset_info: features: - name: input dtype: string - name: output dtype: string - name: embeddings dtype: string splits: - name: train num_bytes: 21718645 num_exam
Hugging Face2024-07-07 更新130
kinianlo/prlang
--- dataset_info: - config_name: conceptnet5_vocabulary_en features: - name: word dtype: string - name: tag dtype: string splits: - name: train num_bytes: 123167929 num_examp
Hugging Face2024-02-08 更新190
cstr/intel_orca_dpo_pairs_de
该数据集是从mayflowergmbh/intel_orca_dpo_pairs_de翻译成德语的,并且被还原为原始的jsonl结构。数据集的语言为德语,使用的许可证为apache-2.0。
Hugging Face2024-03-25 更新130



