LoRAX is an inference server enabling concurrent serving of thousands of fine-tuned LLMs using LoRA adapters without duplicating base model weights.
Open-source Rust-based AI agent framework supporting multiple providers, Model Context Protocol (MCP), plugins, and team coordination.
Official inference and LoRA training package for LTX-2, a generative model for audio–video synthesis. Enables researchers and developers to run and fine-tune the model locally.
OpenLLMetry is an open-source observability framework for LLM and GenAI apps built on OpenTelemetry standards.
Two lightweight frameworks released for efficient LLM serving and scaling, optimized for resource-constrained deployments and production inference.
KRAFTON released Raon-OpenTTS-Pool, a 615K-hour open English speech dataset aggregating 11 sources for training text-to-speech models.
Autonomous research agent that conducts in-depth investigations on any topic using pluggable LLM providers.
MONET is a 104.9M image-text pair dataset curated from 2.9B raw pairs across nine sources, with safety filtering, deduplication, and re-captioning.
Comprehensive GitHub repository with thousands of resources on ethical hacking, bug bounties, DFIR, AI security, vulnerability research, and exploit development.
Metaflow is Netflix's open-source framework for building, managing, and deploying ML systems at scale across cloud platforms (AWS, Azure, GCP).
OpenBMB released UltraData-SFT-2605, a 10M+ example supervised fine-tuning dataset covering math, code, knowledge, and instruction-following used to train MiniCPM5-1B.