My learning notes/codes for ML SYS.

Give us minutes, we give back a faster Mamba. The official implementation of "Faster Vision Mamba is Rebuilt in Minutes via Merged Token Re-training".

This is the official repo for the paper "Accelerating Parallel Sampling of Diffusion Models" Tang et al. ICML 2024 https://openreview.net/forum?id=CjVWen8aJL

PyTorch code for hierarchical k-means -- a data curation method for self-supervised learning

Welcome to the Awesome Feature Learning in Deep Learning Thoery Reading Group! This repository serves as a collaborative platform for scholars, enthusiasts, and anyone interested in delving into th…

depyf is a tool to help you understand and adapt to PyTorch compiler torch.compile.

The official implementation of "2024NeurIPS Dynamic Tuning Towards Parameter and Inference Efficiency for ViT Adaptation"

We introduce a novel approach for parameter generation, named neural network parameter diffusion (p-diff), which employs a standard latent diffusion model to synthesize a new set of parameters

Lossless Training Speed Up by Unbiased Dynamic Data Pruning

Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.

Repo for external large-scale work

The simplest, fastest repository for training/finetuning medium-sized GPTs.

Inference Llama 2 in one file of pure C

基于达摩院在深度学习、计算机视觉、地理空间分析等方向上的技术积累，结合阿里云强大算力支撑，提供多源对地观测数据的云计算分析服务，用数据感知地球世界，让AI助力科学研究。

DAMO-YOLO: a fast and accurate object detection method with some new techs, including NAS backbones, efficient RepGFPN, ZeroHead, AlignedOTA, and distillation enhancement.

A machine learning compiler for GPUs, CPUs, and ML accelerators

A baseline repository of Auto-Parallelism in Training Neural Networks

maximal update parametrization (µP)

The official implementation of ELSA: Enhanced Local Self-Attention for Vision Transformer

Transformer related optimization, including BERT, GPT

Tracking code for the winner of track 1 in the MMP-Tracking Challenge at ICCV 2021 Workshop.

BoxMOT: pluggable SOTA tracking modules for segmentation, object detection and pose estimation models

Single object tracking and segmentation.

[ICCV-2021] TransReID: Transformer-based Object Re-Identification

This is an official implementation for "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows".

Two-stage CenterNet

NanoDet-Plus⚡Super fast and lightweight anchor-free object detection model. 🔥Only 980 KB(int8) / 1.8MB (fp16) and run 97FPS on cellphone🔥

1st Place Solution to ECCV-TAO-2020: Detect and Represent Any Object for Tracking