代码地址：https://github.com/NX-AI/xlstm Hugging Face 地址：https://huggingface.co/NX-AI/xLSTM-7b 通过修改模块架构，研究者优化了吞吐量，在低维空间运行 mLSTM 并添加前馈 MLP 层，同时去除了不必要的组件以提高 GPU 利用率。优化后的架构在保持相似性能的同时，将 token 吞吐量提高了 2 到 4 倍。
Preview Code Blame 330 lines (200 loc) · 38 KB RawExtended Long Short Term Memory (xLSTM): a novel enhancement of the traditional Long Short-Term Memory (LSTM) models The paper introduces xLSTM, an extension of the original LSTM architecture that aims to overcome some of LSTM's limitation...
[tool.poetry] name = "xlstm-torch" version = "0.0.2" description = "xLSTM - Pytorch" license = "MIT" authors = ["Kye Gomez <kye@apac.ai>"] homepage = "https://github.com/kyegomez/xLSTM" documentation = "https://github.com/kyegomez/xLSTM" # Add this if you have documentation...
当人们都以为 Transformer 在语言模型领域稳坐江山的时候,LSTM 又杀回来了 —— 这次,是以 xLSTM 的身份。 5月 8 日,LSTM 提出者和奠基者Sepp Hochreiter在 arXiv 上传了 xLSTM 的预印本论文。 论文的所属机构中还出现了一家叫做「NXAI」的公司,Sepp Hochreiter 表示:「借助 xLSTM,我们缩小了与现有最先进...
Code Edit nx-ai/xlstm official 1,714 smvorwerk/xlstm-cuda 86 gonzalopezgil/xlstm-ts 9 Tasks Edit Language Modeling Language Modelling State Space Models Datasets Edit C4 LAMBADA mC4 Results from the Paper Edit Submit results from this paper to get state-of-the-art GitHub badges ...
models in time series forecasting. PDFAbstract Code AddRemoveMark official mauricekraus/xlstm-mixerofficial 68 Tasks Edit AddRemove Datasets Add Datasetsintroduced or used in this paper Results from the Paper Edit Submitresults from this paperto get state-of-the-art GitHub badges and help the ...
Our team name is DeepLearnAI (CEMRG). The code for this work is available at https://github.com/RespectKnowledge/SSL-based-DINOv2_Vision-LSTM_Head-and-Neck-Tumor_Segmentation.You have full access to this open access chapter, Download conference paper PDF Keywords xLSTM-UNet architecture ...
论文标题:xLSTM 7B: A Recurrent LLM for Fast and Efficient Inference论文地址:https://arxiv.org/pdf/2503.13427代码地址:https://github.com/NX-AI/xlstmHugging Face 地址:https://huggingface.co/NX-AI/xLSTM-7b 通过修改模块架构,研究者优化了吞吐量,在低维空间运行 mLSTM 并添加前馈 MLP 层,同时去除...
Official repository of the xLSTM. Contribute to NX-AI/xlstm development by creating an account on GitHub.
git clone https://github.com/NX-AI/xlstm.gitcdxlstm pip install -e. Requirements This package is based on PyTorch and was tested for versions>=1.8. For a well-tested environment, install theenvironment_pt240cu124.yamlas: conda env create -n xlstm -f environment_pt240cu124.yaml conda ...