Stars
- All languages
- C
- C++
- CMake
- CSS
- CoffeeScript
- Common Workflow Language
- Cuda
- Cython
- Dockerfile
- Fortran
- Go
- HTML
- Haskell
- Java
- JavaScript
- Julia
- Jupyter Notebook
- Lua
- MATLAB
- Nunjucks
- OCaml
- Objective-C
- OpenEdge ABL
- PHP
- Perl
- Protocol Buffer
- PureBasic
- Python
- Ruby
- Rust
- Scala
- Shell
- Swift
- TeX
- TypeScript
- Vim Script
- XSLT
Embodied Reasoning Question Answer (ERQA) Benchmark
NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.
Cosmos-RL is a flexible and scalable Reinforcement Learning framework specialized for Physical AI applications.
Unlocking Lossless Speedups in LLMs via Discrete Diffusion
A lightweight inference engine supporting speculative speculative decoding (SSD).
Jev-compatible API endpoint based on open models (prefill-only)
RL-training knowledge-work agents on APEX-Agents data with SkyRL + Harbor
A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downs…
Memory-bounded compressed sparse attention via streaming top-k. Triton kernels for the DeepSeek-V4 lightning indexer. 32x regime extension on a single H200 | by RightNow https://www.rightnowai.co/
🚀 Ultra Recipe for Training Long-Horizon Search Agents - matching frontier AI's search capability with a 20B model + stateful harness
DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms
A collection of skills for AI financial analysis.
Train transformer language models with reinforcement learning.
deepspeedai / Megatron-DeepSpeed
Forked from NVIDIA/Megatron-LMOngoing research training transformer language models at scale, including: BERT & GPT-2
Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention
gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI
Machine Learning Engineering Open Book
Official PyTorch implementation for "Large Language Diffusion Models"
AlgoTune is a NeurIPS 2025 benchmark made up of 154 math, physics, and computer science problems. The goal is write code that solves each problem, and is faster than existing implementations.
Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents
[NeurIPS 2025 Spotlight] Reasoning Environments for Reinforcement Learning with Verifiable Rewards
A simulation framework for RLHF and alternatives. Develop your RLHF method without collecting human data.
[ICML 2024] Break the Sequential Dependency of LLM Inference Using Lookahead Decoding
Haxophone, an electronic musical instrument that resembles a saxophone
Training and serving large-scale neural networks with auto parallelization.
Shortcuts for Siri using ChatGPT API, supports continuous conversations, configure the API key & save chat records. 由 ChatGPT API 模型驱动的智能 Siri,支持连续对话,配置API key,配置系统prompt,保存聊天记录。