⏰ Collaboratively track deadlines of conferences recommended by CCF (Website, Python Cli, Wechat Applet) / If you find it useful, please star this project, thanks~

ASPLOS'24: Optimal Kernel Orchestration for Tensor Programs with Korch

xLSTM as Generic Vision Backbone

Official code for the paper "Attention as a Hypernetwork"

A mod for GGST that adds training mode features

A multi-level tensor algebra superoptimizer

🏡 Open source home automation that puts local control and privacy first.

MambaOut: Do We Really Need Mamba for Vision?

CUDA Core Compute Libraries

Standalone Flash Attention v2 kernel without libtorch dependency

A simple and efficient Mamba implementation in pure PyTorch and MLX.

An efficient pure-PyTorch implementation of Kolmogorov-Arnold Network (KAN).

Kolmogorov Arnold Networks

TriForce: Lossless Acceleration of Long Sequence Generation with Hierarchical Speculative Decoding

An easy to use PyTorch to TensorRT converter

CUDA/Metal accelerated language model inference

Lightning Attention-2: A Free Lunch for Handling Unlimited Sequence Lengths in Large Language Models

Scripts for fine-tuning Llama2 with composable FSDP & PEFT methods to cover single/multi-node GPUs. Supports default & custom datasets for applications such as summarization & question answering. S…

The code of our paper "InfLLM: Unveiling the Intrinsic Capacity of LLMs for Understanding Extremely Long Sequences with Training-Free Memory"

Implementation of 💍 Ring Attention, from Liu et al. at Berkeley AI, in Pytorch

Token Omission Via Attention

CUDA Kernel Benchmarking Library

"Found in the Middle: How Language Models Use Long Contexts Better via Plug-and-Play Positional Encoding" Zhenyu Zhang, Runjin Chen, Shiwei Liu, Zhewei Yao, Olatunji Ruwase, Beidi Chen, Xiaoxia Wu,…

Easy control for Key-Value Constrained Generative LLM Inference(https://arxiv.org/abs/2402.06262)

Implementation of paper "LM-Infinite: Simple On-the-Fly Length Generalization for Large Language Models"

Running large language models on a single GPU for throughput-oriented scenarios.