Stars
[CVPR 2025] MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis
fastllm是后端无依赖的高性能大模型推理库。同时支持张量并行推理稠密模型和混合模式推理MOE模型，任意10G以上显卡即可推理满血DeepSeek。双路9004/9005服务器+单显卡部署DeepSeek满血满精度原版模型，单并发20tps；INT4量化模型单并发30tps，多并发可达60+。
Distribute and run LLMs with a single file.
LeetCode Training and Evaluation Dataset
FlashMLA: Efficient MLA decoding kernels
OpenCompass is an LLM evaluation platform, supporting a wide range of models (Llama3, Mistral, InternLM2,GPT-4,LLaMa2, Qwen,GLM, Claude, etc) over 100+ datasets.
GPQA: A Graduate-Level Google-Proof Q&A Benchmark
Fully open reproduction of DeepSeek-R1
run DeepSeek-R1 GGUFs on KTransformers
The simplest, fastest repository for training/finetuning medium-sized GPTs.
A Flexible Framework for Experiencing Cutting-edge LLM Inference Optimizations
Apache Arrow is the universal columnar format and multi-language toolbox for fast data interchange and in-memory analytics
Open source platform for the machine learning lifecycle
Sacred is a tool to help you configure, organize, log and reproduce experiments developed at IDSIA.
High-speed download of LLaMA, Facebook's 65B parameter GPT model
Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.
Pretrain, finetune ANY AI model of ANY size on multiple GPUs, TPUs with zero code changes.
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
CUDA integration for Python, plus shiny features
Code from various chapters in OSTEP (http://www.ostep.org)