dsgiitr / kge-clip

Knowledge Graph Embeddings (KGE) for RAG-LLMs. Our goal was to compare the mathematical differences between Traditional Static Multimodal Vector Embeddings (TVE) from Word2Vec and CLIP encoders for {text:image} datasets, and Knowledge Graph Embeddings (KGE) generated with REBEL triplets trained on PyKeen.
12Updated last month

Related projects

Alternatives and complementary repositories for kge-clip