Fine-tuning an LLM used to mean writing 400 lines of Python just to get started. We got tired of watching domain experts give up before the first training run. So we built a LynxKite LLM Training plugin: a visual, drag-and-drop workflow for the entire fine-tuning process. Load a model, wire up your dataset, define reward functions, train with supervised fine-tuning or GRPO reinforcement learning — all as connected nodes on a canvas.

LynxKite has a flexible graph neural network designer where you can choose from a variety of GNN layers. Graph attention networks are a great choice in many applications. In addition to providing good predictions, GATs are unique in that the attention mechanism gives us insights into what the prediction is based on. This post will explore how we can interrogate the inner workings of a graph neural network and generate a human-readable explanation for its predictions.