nops · PyPI

Neural Operators made simple.

graph LR
    subgraph nops_FNO [nops: FNO Architecture]
        direction LR
        Input["Input<br/>(Batch, C_in, X, Y...)"] --> Grid{Append<br/>Grid?}
        Grid -- Yes --> InputGrid[Input + Grid]
        Grid -- No --> InputGrid
        
        InputGrid --> Lifting["Lifting Layer<br/>(Projection P)"]
        
        subgraph FourierLayers [Fourier Layers x N]
            direction TB
            FFT[FFT] --> SpecConv["Spectral Conv<br/>(R * Weights)"]
            SpecConv --> iFFT["Inverse FFT"]
            iFFT --> Skip{"Skip<br/>Connection"}
            
            Skip --> Attention{"Attention<br/>Gating?"}
            Attention -- Yes --> AttnBlock["Attention Weights"]
            Attention -- No --> SumBlock["Sum"]
            
            AttnBlock --> OutputLayer
            SumBlock --> OutputLayer
        end
        
        Lifting --> FourierLayers
        FourierLayers --> Projection["Projection Layer<br/>(Q)"]
        Projection --> Output["Output<br/>(Batch, C_out, X, Y...)"]
    end

A human-centric interface for operating with Neural Operators (FNO, GNO, DeepONet, etc.).

nops is designed to be an elegant, high-level layer over complex neural operator architectures. Whether you are a researcher prototyping a new PDE solver or an engineer deploying a model to production, nops abstracts the complexity while retaining the flexibility you need.

✨ Features

Human-Friendly API: Intuitive classes and methods that make sense.
Multi-Dimensional: Seamless support for 1D, 2D, and 3D FNO models.
Production Ready: Built with modularity, type hinting, and efficiency in mind.
Batteries Included:
Attention Gating: Advanced mechanisms to weight parallel Fourier branches.
Flexible Gridding: Automatic handling of grid coordinates.
Padding Management: Easy handling of boundary conditions.

📦 Installation

Install nops easily via pip:

pip install nops

Note: Requires Python 3.12+

🚀 Quick Start

Solving a PDE with a Fourier Neural Operator (FNO) has never been this simple.

import torch
import torch.nn as nn
from nops.fno.models.original import FNO

# 1. Define your configuration
model = FNO(
    modes=[16, 16],               # 16 Fourier modes for height and width
    num_fourier_layers=4,         # Depth of the network
    in_channels=3,                # (e.g., u, v, p)
    out_channels=1,               # (e.g., pressure field)
    lifting_channels=64,
    projection_channels=64,
    mid_channels=64,
    activation=nn.GELU(),
    add_grid=True,                # Auto-append x,y coordinates
    attn_gating=True              # Enable attention mechanism
)

# 2. Forward pass
# Batch size: 8, Channels: 3, Resolution: 128x128
input_data = torch.randn(8, 3, 128, 128)
output = model(input_data)

print(f"Output Field Shape: {output.shape}")
# > Output Field Shape: torch.Size([8, 1, 128, 128])

🗺️ Roadmap

FNO (Fourier Neural Operator): Original implementation with attention extensions.
DeepONet: Implementation of Deep Operator Networks.
GNO (Graph Neural Operator): Support for irregular meshes.
Physics-Informed Losses: Built-in loss functions for PDE constraints.
Data Loaders: Standard datasets (Burgers, Navier-Stokes, Darcy Flow).

🤝 Contributing

We welcome contributions! nops is a community-driven project to make neural operators accessible to everyone.

Fork the repository.
Create a Feature Branch (git checkout -b feature/AmazingFeature).
Commit your changes (git commit -m 'Add some AmazingFeature').
Push to the branch (git push origin feature/AmazingFeature).
Open a Pull Request.

Please ensure your code follows the existing style and includes tests.

📄 License

Distributed under the MIT License. See LICENSE for more information.

_{Built with ❤️ by the nops community.}

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nops-0.3.0.tar.gz (505.4 kB view details)

Uploaded Apr 19, 2026 Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

nops-0.3.0-py3-none-any.whl (96.9 kB view details)

Uploaded Apr 19, 2026 Python 3

File details

Details for the file nops-0.3.0.tar.gz.

File metadata

Download URL: nops-0.3.0.tar.gz
Upload date: Apr 19, 2026
Size: 505.4 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for nops-0.3.0.tar.gz
Algorithm	Hash digest
SHA256	`1cf6858fe205758fec7306ca1e346ad4633fc3d5fc11a02c895aac461534f6e9`
MD5	`bb5aea0bf693d802ddf98094b675a0e1`
BLAKE2b-256	`ce887b6751970cd3e4a19f01a1e5b07782fbdcce8b84949654156648702cf376`

See more details on using hashes here.

Provenance

The following attestation bundles were made for nops-0.3.0.tar.gz:

Publisher: pypi.yml on abelsr/nops

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Statement:
- Statement type: https://in-toto.io/Statement/v1
- Predicate type: https://docs.pypi.org/attestations/publish/v1
- Subject name: nops-0.3.0.tar.gz
- Subject digest: 1cf6858fe205758fec7306ca1e346ad4633fc3d5fc11a02c895aac461534f6e9
- Sigstore transparency entry: 1340216465
- Sigstore integration time: Apr 19, 2026
Source repository:
- Permalink: abelsr/nops@b56fbfbaa7cf8d1b6e5634abf004e17e0e6f862a
- Branch / Tag: refs/tags/v0.3.0
- Owner: https://github.com/abelsr
- Access: public
Publication detail:
- Token Issuer: https://token.actions.githubusercontent.com
- Runner Environment: github-hosted
- Publication workflow: pypi.yml@b56fbfbaa7cf8d1b6e5634abf004e17e0e6f862a
- Trigger Event: release