NVIDIA Corporation
- 32.1k followers
- 2788 San Tomas Expressway, Santa Clara, CA, 95051
- https://nvidia.com
Pinned Loading
Repositories
-
- Model-Optimizer Public
A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.
-
- Nomos Public
Stops AI agents from taking unwanted or overeager actions. Nomos reviews each proposed action against the agent's own context (its permissions, task, justification, blast radius and signed receipts of what it has observed) using OPA policy and an LLM council. Approved actions get a signed, single-use Writ; the executor runs exactly that.
- physicsnemo-serve Public
PhysicsNeMo-Serve is a Rust-based inference serving package for MLOps engineers that turns supported PhysicsNeMo inference pipelines into production-ready GPU services, delivering a shorter checkpoint-to-endpoint path with backend-aware deployment support
- kumo-engine Public
Kumo Engine is an accelerator for context generation over structured data, designed for teams building and running predictive models in production at scale.
- cudnn-frontend Public
cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.