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

Itransformer

Itransformer is a public Python implementation of the iTransformer attention model for time-series forecasting. It includes a standard model and repository examples for variants that add attention over time tokens or Fourier-derived tokens.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Itransformer — 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 the repository when you want to construct an iTransformer model in Python and work with time-series tensors in your own workflow, rather than use a hosted forecasting service.

Who it is for

  • Python developers prototyping time-series forecasting
  • Researchers exploring attention-based forecasting models
  • Practitioners comparing forecasting approaches within a data-analysis workflow

Documented uses

  • Generate forecasts from multivariate time-series inputs at one or more prediction lengths.
  • Prototype attention-based forecasting experiments in Python.
  • Explore variants using time-token or Fourier-token inputs.
Before you choose

Requirements & limits

  • Installation is documented through pip, and the usage examples import PyTorch.
  • The demonstrated input has the shape batch, lookback length and variates; outputs are grouped by prediction length.
  • The repository labels itself an unofficial implementation and links to a separate official implementation.
  • The README marks multiple tokens per variate as experimental and describes the Fourier-token model in an Experimental section.

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

Install the package and inspect the forecasting example

Set up the documented Python package and follow the README’s model-and-input example structure.

Before you start

  • Python environment
  • The iTransformer package, installed with pip
  • PyTorch, imported in the documented example
  • Time-series input shaped as batch, lookback length, and variates

Steps

  1. Install the package with the command below.

    Documented command
    pip install iTransformer

  2. Import torch and iTransformer in Python.

  3. Construct the model with the documented variate count, lookback length, model dimensions and prediction lengths.

  4. Pass a time-series tensor to the model and inspect the returned predictions, organized by prediction length.

Expected outcome

A Python call pattern that returns prediction tensors for the configured forecast horizons.

Guide limitations

  • The README’s example uses randomly generated data; it does not document a dataset-preparation workflow.
  • This is an unofficial implementation, and the repository points to a separate official implementation.
  • The example documents a usage pattern, not a verified forecast result or model performance guarantee.
Documentation 1 ↗
Evidence and transparency

Sources & review

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