cisco-ai/model-provenance-kit

该数据集分发预计算的深度信号权重指纹，用于与模型来源工具包（Model Provenance Kit）一起使用。主要交付物是`deep-signals.zip`：一个压缩的Apache Parquet文件存档，存储从Hugging Face Hub或类似来源公开发布的Transformer（及相关）模型权重中提取的密集数值特征。每个文件以长形式表格模式编码多信号权重级证据（嵌入锚点、规范层向量、层能量剖面、嵌入规范直方图和每层相关性样式摘要）。这些指纹支持快速相似性搜索和来源匹配。数据集不包含自然语言文本，也不适用于通常意义上的语言模型预训练。

This repository distributes pre-computed deep-signal weight fingerprints for use with Model Provenance Kit. The primary deliverable is `deep-signals.zip`: a compressed archive of Apache Parquet files that store dense numerical features extracted from publicly released transformer (and related) model weights on the Hugging Face Hub or equivalent sources. Each file encodes multi-signal weight-level evidence (embedding anchors, norm-layer vectors, layer energy profiles, embedding norm histograms, and per-layer correlation-style summaries) in a long-form tabular schema. These fingerprints enable fast similarity search and provenance-style matching. They are not natural-language text and are not intended for language-model pretraining in the usual sense.