# Lrnnx

## Docs

- [Classifier](https://mintlify.wiki/SforAiDl/lrnnx/api/architectures/classifier.md): Sequence classifier and regressor with LRNN backbone
- [Embedding Classes](https://mintlify.wiki/SforAiDl/lrnnx/api/architectures/embedding.md): Token and positional embedding modules for sequence models
- [LRNNLMHeadModel](https://mintlify.wiki/SforAiDl/lrnnx/api/architectures/language-model.md): LRNN Language Model with causal language modeling head
- [LRU_UNet](https://mintlify.wiki/SforAiDl/lrnnx/api/architectures/unet.md): Linear Recurrent Unit (LRU) based U-Net for sequence tasks
- [LRNN Base Class](https://mintlify.wiki/SforAiDl/lrnnx/api/core/base.md): Base model class for LRNN with abstract forward method and discretization support
- [Convolution Operations](https://mintlify.wiki/SforAiDl/lrnnx/api/core/convolution.md): FFT-based convolution operations with optimized einsum contractions for efficient sequence modeling
- [Discretization Functions](https://mintlify.wiki/SforAiDl/lrnnx/api/core/discretization.md): Methods for converting continuous-time state space models to discrete-time representations
- [Block](https://mintlify.wiki/SforAiDl/lrnnx/api/layers/block.md): Block layer that wraps mixer modules with normalization and residual connections
- [MHA (Multi-Head Attention)](https://mintlify.wiki/SforAiDl/lrnnx/api/layers/mha.md): Multi-head self-attention with optional convolution, rotary embeddings, and integrated MLP
- [MLP Layers](https://mintlify.wiki/SforAiDl/lrnnx/api/layers/mlp.md): Multi-layer perceptron implementations including gated variants
- [LTI_LRNN](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/base.md): Base class for Linear Time-Invariant LRNN models
- [Centaurus](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/centaurus.md): Multi-mode State Space model with intra-state mixing
- [LRU](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/lru.md): Linear Recurrent Unit with learned complex diagonal dynamics
- [S4](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/s4.md): Structured State Space Sequence model with DPLR parameterization
- [S4D](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/s4d.md): Structured State Space Sequence model with diagonal parameterization
- [S5](https://mintlify.wiki/SforAiDl/lrnnx/api/lti/s5.md): Simplified State Space model with complex diagonal parameterization
- [LTV_LRNN](https://mintlify.wiki/SforAiDl/lrnnx/api/ltv/base.md): Base class for Linear Time-Varying LRNN models
- [Mamba](https://mintlify.wiki/SforAiDl/lrnnx/api/ltv/mamba.md): Selective State Space Model with event-based processing support
- [RGLRU](https://mintlify.wiki/SforAiDl/lrnnx/api/ltv/rglru.md): Recurrent Gated Linear Recurrent Unit from Griffin architecture
- [S7](https://mintlify.wiki/SforAiDl/lrnnx/api/ltv/s7.md): Selective and Simplified State Space Layer for sequence modeling
- [S4 Kernel Operations](https://mintlify.wiki/SforAiDl/lrnnx/api/ops/s4-kernel.md): S4 kernel operations for diagonal and low-rank state space models
- [Selective Scan Operations](https://mintlify.wiki/SforAiDl/lrnnx/api/ops/selective-scan.md): CUDA-accelerated selective scan operations for Mamba SSM models
- [Simplified Scan Operations](https://mintlify.wiki/SforAiDl/lrnnx/api/ops/simplified-scan.md): CUDA-accelerated simplified scan operations for S5-style SSM models
- [Hierarchical Classifier](https://mintlify.wiki/SforAiDl/lrnnx/architectures/classifier.md): LRNN-based classifier for sequence classification and regression tasks
- [Language Model](https://mintlify.wiki/SforAiDl/lrnnx/architectures/language-model.md): LRNN-based language model architecture for autoregressive text generation
- [U-Net](https://mintlify.wiki/SforAiDl/lrnnx/architectures/unet.md): LRU-based U-Net architecture for sequence-to-sequence tasks
- [Discretization methods](https://mintlify.wiki/SforAiDl/lrnnx/concepts/discretization.md): Converting continuous-time state-space models to discrete-time for implementation
- [Linear RNNs](https://mintlify.wiki/SforAiDl/lrnnx/concepts/linear-rnns.md): Understanding linear recurrent neural networks and why they matter for sequence modeling
- [LTI vs LTV models](https://mintlify.wiki/SforAiDl/lrnnx/concepts/lti-vs-ltv.md): Understanding the difference between Linear Time-Invariant and Linear Time-Varying models
- [Custom CUDA Kernels](https://mintlify.wiki/SforAiDl/lrnnx/guides/custom-kernels.md): High-performance CUDA kernels for selective scan, simplified scan, and structured operations
- [Inference Guide](https://mintlify.wiki/SforAiDl/lrnnx/guides/inference.md): Fast autoregressive generation with CUDA graphs for 10x speedup
- [Training Guide](https://mintlify.wiki/SforAiDl/lrnnx/guides/training.md): Learn how to train lrnnx models with forward and backward passes
- [Installation](https://mintlify.wiki/SforAiDl/lrnnx/installation.md): Install lrnnx for your PyTorch projects
- [Lrnnx documentation](https://mintlify.wiki/SforAiDl/lrnnx/introduction.md): A unified PyTorch library for Linear RNN architectures
- [Centaurus](https://mintlify.wiki/SforAiDl/lrnnx/models/lti/centaurus.md): Multi-mode state space model with intra-state mixing
- [LRU](https://mintlify.wiki/SforAiDl/lrnnx/models/lti/lru.md): Linear Recurrent Unit with diagonal complex dynamics
- [S4](https://mintlify.wiki/SforAiDl/lrnnx/models/lti/s4.md): Structured State Space Sequence model with DPLR parameterization
- [S4D](https://mintlify.wiki/SforAiDl/lrnnx/models/lti/s4d.md): Diagonal variant of S4 with simplified parameterization
- [S5](https://mintlify.wiki/SforAiDl/lrnnx/models/lti/s5.md): Simplified State Space model with multiple discretization options
- [Mamba](https://mintlify.wiki/SforAiDl/lrnnx/models/ltv/mamba.md): Selective state space model with input-dependent dynamics
- [RG-LRU](https://mintlify.wiki/SforAiDl/lrnnx/models/ltv/rglru.md): Recurrent Gated Linear Recurrent Unit from Griffin architecture
- [S7](https://mintlify.wiki/SforAiDl/lrnnx/models/ltv/s7.md): Selective and Simplified State Space model with HiPPO initialization
- [Models Overview](https://mintlify.wiki/SforAiDl/lrnnx/models/overview.md): Overview of all state space models in lrnnx
- [Quickstart](https://mintlify.wiki/SforAiDl/lrnnx/quickstart.md): Get started with lrnnx in minutes
- [Hierarchical Classification with LRNN](https://mintlify.wiki/SforAiDl/lrnnx/tutorials/hierarchical-classification.md): Train a hierarchical classifier on ListOps using LRU and S5 models
- [U-Net Audio Denoising](https://mintlify.wiki/SforAiDl/lrnnx/tutorials/unet-denoising.md): Build a 1D U-Net audio denoiser using LRU for sequence modeling