M-Chimiste/Llama-3-8B-RDF-Experiment
M-Chimiste/Llama-3-8B-RDF-Experiment is an 8 billion parameter experimental model, fine-tuned from NousResearch/Hermes-2-Pro-Llama-3-8B. This model is specifically designed to construct knowledge graph triples, leveraging its Llama-3 architecture and an 8192-token context length. It specializes in knowledge graph annotation, responding in a structured JSON format according to a predefined schema.
Loading preview...
Model Overview
M-Chimiste/Llama-3-8B-RDF-Experiment is an 8 billion parameter language model, fine-tuned from the NousResearch/Hermes-2-Pro-Llama-3-8B base. This model is an experimental finetune specifically developed to explore its capability in constructing knowledge graph triples.
Key Capabilities
- Knowledge Graph Annotation: The primary function of this model is to act as an expert knowledge graph annotator.
- Structured JSON Output: It is designed to respond exclusively in a JSON format, adhering to a specified schema for representing triples (subject, predicate, object).
- Llama-3 Architecture: Built upon the Llama-3 8B foundation, providing a robust base for language understanding and generation.
- qLoRA Finetuning: The model was finetuned using qLoRA on Unsloth, then merged back to a 16-bit format, indicating an efficient training approach.
Recommended Use
This model is particularly well-suited for tasks requiring the extraction and structuring of information into RDF-like knowledge graph triples. Users should utilize the apply_chat_template feature with the provided system prompt to ensure correct JSON output for knowledge graph annotation. It is ideal for researchers and developers experimenting with automated knowledge graph construction from text.