MLRun is a high-performance, open-source ML experiment tracking platform designed for modern AI workflows. It provides infrastructure for logging metrics, parameters, and artifacts from training runs, with built-in support for LLM evaluations, agent tracing, and high-frequency data ingestion.
GitHubGhostVis is a vision-language model built on top of nanochat, transforming a text-only LLM into a full multimodal system capable of understanding both images and text. Like nanochat, it maintains a clean, minimal, hackable codebase designed to run on a single 8xH100 node. From pretraining / mid / vision alignment with CLIP, Rl, and Slang inference, it is an end to end pipeline, and with swiglu architecture + fused triton kernels for the LLM text architecture.
GitHub