$\texttt{lrnnx}$: A library for Linear RNNs
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910016666075136 |
|---|---|
| author | Bania, Karan Kalburgi, Soham Tanwar, Manit Dhruthi Nagarsekar, Aditya Mestha, Harshvardhan Chibber, Naman Deshmukh, Raj Sathyanarayanan, Anish Rathore, Aarush Chheda, Pratham |
| author_facet | Bania, Karan Kalburgi, Soham Tanwar, Manit Dhruthi Nagarsekar, Aditya Mestha, Harshvardhan Chibber, Naman Deshmukh, Raj Sathyanarayanan, Anish Rathore, Aarush Chheda, Pratham |
| contents | Linear recurrent neural networks (LRNNs) provide a structured approach to sequence modeling that bridges classical linear dynamical systems and modern deep learning, offering both expressive power and theoretical guarantees on stability and trainability. In recent years, multiple LRNN-based architectures have been proposed, each introducing distinct parameterizations, discretization schemes, and implementation constraints. However, existing implementations are fragmented across different software frameworks, often rely on framework-specific optimizations, and in some cases require custom CUDA kernels or lack publicly available code altogether. As a result, using, comparing, or extending LRNNs requires substantial implementation effort. To address this, we introduce $\texttt{lrnnx}$, a unified software library that implements several modern LRNN architectures under a common interface. The library exposes multiple levels of control, allowing users to work directly with core components or higher-level model abstractions. $\texttt{lrnnx}$ aims to improve accessibility, reproducibility, and extensibility of LRNN research and applications. We make our code available under a permissive MIT license. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_08810 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | $\texttt{lrnnx}$: A library for Linear RNNs Bania, Karan Kalburgi, Soham Tanwar, Manit Dhruthi Nagarsekar, Aditya Mestha, Harshvardhan Chibber, Naman Deshmukh, Raj Sathyanarayanan, Anish Rathore, Aarush Chheda, Pratham Machine Learning Artificial Intelligence Linear recurrent neural networks (LRNNs) provide a structured approach to sequence modeling that bridges classical linear dynamical systems and modern deep learning, offering both expressive power and theoretical guarantees on stability and trainability. In recent years, multiple LRNN-based architectures have been proposed, each introducing distinct parameterizations, discretization schemes, and implementation constraints. However, existing implementations are fragmented across different software frameworks, often rely on framework-specific optimizations, and in some cases require custom CUDA kernels or lack publicly available code altogether. As a result, using, comparing, or extending LRNNs requires substantial implementation effort. To address this, we introduce $\texttt{lrnnx}$, a unified software library that implements several modern LRNN architectures under a common interface. The library exposes multiple levels of control, allowing users to work directly with core components or higher-level model abstractions. $\texttt{lrnnx}$ aims to improve accessibility, reproducibility, and extensibility of LRNN research and applications. We make our code available under a permissive MIT license. |
| title | $\texttt{lrnnx}$: A library for Linear RNNs |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2602.08810 |