NeuralOperator
NeuralOperator is the reference implementation library for neural operators, providing Fourier Neural Operator (FNO), DeepONet and other mainstream architectures with training pipelines under the MIT license (free for commercial use and modification). It standardizes learning 'function-to-function' maps and is the core tool for building physical-field surrogates (replacing CFD solves in milliseconds).
Use cases: Training fast surrogates for flow / temperature fields; replacing slow numerical solves for design iteration.
Tags: Neural OperatorFNODeepONetMIT
Related entries
- PhysicsNeMo · NVIDIA's physics-ML training framework (formerly Modulus) with PINN, neural oper
- POSEIDON · ETH Zurich's PDE foundation model and downstream fine-tuning suite, using semigr
- NeuralOperator v2.0.0 Release · NeuralOperator 2.0.0 major release introduces Codano model, improves documentati
- AIPerf: LLM Inference Benchmarking at Scale · AIPerf is an LLM inference benchmarking tool from NVIDIA designed to evaluate th
- HiDream-O1-Video by ZhiXiang · HiDream-O1-Video-1.0 is a video generation model that demonstrates strong unders