OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Omninet Pytorch

Omninet Pytorch is a Python/PyTorch library implementing OmniNet transformer attention, pooling tokens from earlier layers so they can be attended to across the model. It exposes `Omninet` and `OmninetCausal` modules for machine-learning experiments.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Omninet Pytorch — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when an existing PyTorch project needs an implementation to explore attention across tokens from earlier layers, or to compare configuration choices in a transformer prototype.

Who it is for

  • PyTorch developers prototyping transformer architectures with cross-layer attention.
  • Researchers comparing attention configurations such as model depth, head dimensions and token-pooling interval.

Documented uses

  • Configure an OmniNet module and pass it a tensor and mask.
  • Experiment with attention over tokens from previous transformer layers.
  • Explore a causal module while accounting for its stated difference from the paper.
Before you choose

Requirements & limits

The repository is public and specifies the MIT license. The README documents installation through pip.

  • The standard Omninet example takes a PyTorch tensor and mask and returns an output with the same shape.
  • The README describes pooling tokens every N layers as central to attending across tokens from multiple layers.
  • OmninetCausal is explicitly described as not faithful to the paper; it currently uses layer axial attention with layer positional embeddings.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up and configure an Omninet module

Install the package and prepare the documented non-causal module call with a tensor and mask.

Before you start

  • Python environment with pip
  • PyTorch

Steps

  1. Install the package with the command below.

    Documented command
    pip install omninet-pytorch

  2. Import torch and Omninet from omninet_pytorch.

  3. Construct Omninet with model dimension, depth, head dimensions, head count, pooling interval and the documented dropout settings.

  4. Prepare a PyTorch input tensor and Boolean mask, then pass the input to the module with mask=mask.

Expected outcome

The README example shows an output tensor with shape `(1, 1024, 512)` for an input of that shape.

Guide limitations

  • The example illustrates a configured call; it does not establish benchmark performance or suitability for a particular task.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
lucidrains/omninet-pytorch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction