ParaRNN: Unlocking Parallel Training of Nonlinear RNNs for Large Language Models

TL;DR

ParaRNN introduces a parallel training framework for nonlinear RNNs, overcoming the sequential bottleneck and enabling large-scale training with significant speedups, matching the performance of Transformers.

Contribution

It presents a novel parallelization method for nonlinear RNNs, allowing efficient training at large scales and broadening the scope of sequence modeling architectures.

Findings

Achieved up to 665x speedup over naive sequential methods.

Successfully trained 7B parameter nonlinear RNNs with competitive perplexity.

Open-sourced the ParaRNN framework for community use.

Abstract

Recurrent Neural Networks (RNNs) laid the foundation for sequence modeling, but their intrinsic sequential nature restricts parallel computation, creating a fundamental barrier to scaling. This has led to the dominance of parallelizable architectures like Transformers and, more recently, State Space Models (SSMs). While SSMs achieve efficient parallelization through structured linear recurrences, this linearity constraint limits their expressive power and precludes modeling complex, nonlinear sequence-wise dependencies. To address this, we present ParaRNN, a framework that breaks the sequence-parallelization barrier for nonlinear RNNs. Building on prior work, we cast the sequence of nonlinear recurrence relationships as a single system of equations, which we solve in parallel using Newton's iterations combined with custom parallel reductions. Our implementation achieves speedups of up to 665x over naive sequential application, allowing training nonlinear RNNs at unprecedented scales. To showcase this, we apply ParaRNN to adaptations of LSTM and GRU architectures, successfully training models of 7B parameters that attain perplexity comparable to similarly-sized Transformers and Mamba2 architectures. To accelerate research in efficient sequence modeling, we release the ParaRNN codebase as an open-source framework for automatic training-parallelization of nonlinear RNNs, enabling researchers and practitioners to explore new nonlinear RNN models at scale.