🚀 Awesome System for Machine Learning ⚡️ AI System Papers and Industry Practice. ⚡️ System for Machine Learning, LLM (Large Language Model), GenAI (Generative AI). 🍻 OSDI, NSDI, SIGCOMM, SoCC, MLSys, etc. 🗃️ Llama3, Mistral, etc. 🧑‍💻 Video Tutorials.

🔥🔥High-Performance Face Recognition Library on PaddlePaddle & PyTorch🔥🔥

Implementation of Dreambooth (https://arxiv.org/abs/2208.12242) by way of Textual Inversion (https://arxiv.org/abs/2208.01618) for Stable Diffusion (https://arxiv.org/abs/2112.10752). Tweaks focused on training faces, objects, and styles.

Resource scheduling and cluster management for AI

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

Data-centric LLM training with dynamic sample selection, domain mixture optimization, and example reweighting inside the LLaMA-Factory training loop.

AI-Native & Cloud-Native FS: A high-performance file semantic layer for cloud object storage, integrated with high-speed cache

SkyChat是一款基于中文GPT-3 api的聊天机器人项目。它可以像chatGPT一样，实现人机聊天、问答、中英文互译、对对联、写古诗等任务。| SkyChat is a Chatbot project based on Chinese GPT3 API. Like chatGPT, it can do human-machine chat, question and answer, and can also complete tasks such as Chinese-English or English-Chinese translation, content continuation, couplets, and Chinese ancient poems writing.

Run Slurm in Kubernetes

SkyText是由奇点智源发布的中文GPT3预训练大模型，可以进行文章续写、对话、中英翻译、内容风格生成、推理、诗词对联等不同任务。| SkyText is a Chinese GPT3 pre-trained large model released by Singularity-AI, which can perform different tasks such as chatting, Q&A, and Chinese-English translation.

A complete end-to-end demonstration in which we collect training data in Unity and use that data to train a deep neural network to predict the pose of a cube. This model is then deployed in a simulated robotic pick-and-place task.

Stable Diffusion Regularization Images in 512px, 768px and 1024px on 1.5, 2.1 and SDXL 1.0 checkpoints

Open-source PyTorch tool that traces live training signals back to the data samples causing them. Pause the training to find mislabels, outliers, and class imbalance in image, video, and LiDAR datasets. And train again.

PyTorch model training and layer saturation monitor

A script that can be used to capture various porn novels for machine learning / 一个可以用于抓取各类色情小说用于机器学习的脚本

Accelerating AI Training and Inference from Storage Perspective (Must-read Papers on Storage for AI)

Give your computer an AI Brain

(Google Play Store App) Pipeline to convert real-life chess boards into 2D digital format (FEN) from images and live camera feeds. The system has two versions: a real-time pipeline using OAK-D Lite and a high-precision static-image pipeline. RF-DETR and YOLO is used to detect chess pieces, and traditional cv methods determine board positioning.

A ready-to-use notebook!

Extension for CSGO game: a real-time object detection extension using a trained YOLOv8 model with TensorRT-accelerated inference to enhance aiming precision within Counter-Strike: Global Offensive (CSGO).