Image ranker via click-to-choose using locally running UIan idea by jeremy_moffitt DescriptionCreate an locally deployable application that allows the user to repeatedly choose between two images from a set to rank them over the course of multiple comparisons. The image set should be loadable from a local file folder the user specifies, then the program displays 2 images at a time to the user who selects their preferred image, random comparisons continue until the user decides they are done, or some threshold for comparisons is met (for example, all images except 1 have "lost" 5 times). The resulting stack sort is available to the user. As a stretch goal the images could be renamed or copy+rename based on the results. Additional stretch goal would be to sort images from a library of URLs (or perhaps a flickr stream). An unlikely but ambitious additional stretch goal is to filter out blurry images using some sort of AI plugin. |
Building and Distilling a Custom LLM Using Rancher Dataa project by rcabello DescriptionThis project explores how to create a custom Large Language Model (LLM) trained specifically on Rancher related data. We will explore LLM distillation techniques to produce a smaller, more efficient model suitable for deployment in resource-constrained environments. We will also explore how Retrieval-Augmented Generation (RAG) can enhance model performance by combining the strengths of fine-tuning with dynamic knowledge retrieval. 
|