rntc/bb-tt-3-pretrain

Biomed-FR-v3 高质量预训练数据集包含法语生物医药文本，使用 rntc/biomed-fr-v2-classifier 模型进行了20个不同分类和回归任务的注释。数据集共有2,782,686个样本，41个列，涵盖25个注释任务，专注于生物医药/临床领域。数据集经过过滤，以确保94.6%的数据适合预训练。它包括完整的注释覆盖范围、质量度量、临床重点和正确的列顺序等关键特性。

The Biomed-FR-v3 High-Quality Pretraining Dataset contains French biomedical text annotated with 20 different classification and regression tasks using the `rntc/biomed-fr-v2-classifier` model. The dataset comprises 2,782,686 samples across 41 columns, covering 25 annotation tasks, focusing on the biomedical and clinical domain. The data has been filtered to ensure 94.6% of it is suitable for pretraining. Key features include complete annotation coverage, quality metrics, clinical focus, and proper column order.