Lists (1)
Stars
Reverse engineer anything with agents, from app behavior down to native binaries.
AI 基础知识 - GPU 架构、CUDA 编程、大模型基础及AI Agent 相关知识。
DeepInfer: High-performance, large-scale inference framework for DeepSeek models
🧠 Train a 64M-parameter LLM from scratch in just 2h!
Mixture-of-experts (MoE) training megakernel for NVL72s
NVIDIA curated collection of educational resources related to general purpose GPU programming.
My learning notes for ML SYS.
A kernel library written in tilelang
INFOCOM'26-Vedrfolnir: RDMA Network Performance Anomalies Diagnosis in Collective Communications
OpenDiLoCo: An Open-Source Framework for Globally Distributed Low-Communication Training
MSCCL++: A GPU-driven communication stack for scalable AI applications
ns3-gym - The Playground for Reinforcement Learning in Networking Research
Training materials associated with NVIDIA's CUDA Training Series (www.olcf.ornl.gov/cuda-training-series/)
bytedance / ps-lite
Forked from dmlc/ps-liteA lightweight parameter server interface
Venus Collective Communication Library, supported by SII and Infrawaves.
The official implementation for the intra-stage fusion technique introduced in https://arxiv.org/abs/2409.13221
Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.
The simplest, fastest repository for training/finetuning small-sized VLMs.
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…
slime is an LLM training framework for RL Scaling.
Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.
Enable the interaction between ns-3 and popular frameworks using Python, which mean you can train and test your AI algorithms in ns-3 without changing any frameworks you are using now!