thuml/iTransformer
Inverted transformer architecture for multivariate time series forecasting tasks.
Not currently ranked — collecting fresh signals.
star history
iTransformer rethinks the transformer architecture by inverting the attention mechanism to focus on temporal variables rather than time steps, achieving state-of-the-art results on major time series forecasting benchmarks. The implementation provides training, evaluation, and inference code along with pre-trained models, and has been adopted into production ML frameworks like GluonTS and NeuralForecast.
Frequently asked
- What is thuml/iTransformer?
- Inverted transformer architecture for multivariate time series forecasting tasks.
- Is iTransformer open source?
- Yes — thuml/iTransformer is open source, released under the MIT license.
- What language is iTransformer written in?
- thuml/iTransformer is primarily written in Python.
- How popular is iTransformer?
- thuml/iTransformer has 2.2k stars on GitHub.
- Where can I find iTransformer?
- thuml/iTransformer is on GitHub at https://github.com/thuml/iTransformer.