A construction kit for reinforcement learning environment management.

This is a course on Deep Learning-based Recommender Systems taught at HSE University, academic year 2025/26.

The batteries-included agent harness.

Developer Asset Hub for NVIDIA Nemotron — A one-stop resource for training recipes, usage cookbooks, datasets, and full end-to-end reference examples to build with Nemotron models

Material for gpu-mode lectures

🚀 Efficient implementations for emerging model architectures

DeeperGEMM: crazy optimized version

FlashInfer: Kernel Library for LLM Serving

A high-throughput and memory-efficient inference and serving engine for LLMs

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

Code for "Learning to summarize from human feedback"

CLI interfaces & config objects, from types

Agent Reinforcement Trainer: train multi-step agents for real-world tasks using GRPO. Give your agents on-the-job training. Reinforcement learning for Qwen3.6, GPT-OSS, Llama, and more!

Parallelized search for matrix multiplication schemes using flip graphs on PyTorch

A collection of notebooks/recipes showcasing usecases of open-source models with Together AI.

DeepEP: an efficient expert-parallel communication library

A Datacenter Scale Distributed Inference Serving Framework

Official inference framework for 1-bit LLMs

Optimized primitives for collective multi-GPU communication

Collective communications library with various primitives for multi-machine training.

ContextualAI's text-to-SQL pipeline for BIRD benchmark

Efficient Triton Kernels for LLM Training

Everything about the SmolLM and SmolVLM family of models

Complete solutions to the Programming Massively Parallel Processors Edition 4

🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support

gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI

Code and training scripts for FlexOlmo

slime is an LLM post-training framework for RL Scaling.

ZeroSearch: Incentivize the Search Capability of LLMs without Searching