PETSA vs multiresolution-time-series-transformer
Time-series forecasting projects compared
PETSA adapts pre-trained time-series forecasters at inference time using compact calibration modules. multiresolution-time-series-transformer is a PyTorch forecasting model that processes inputs at several temporal scales; the first focuses on adaptation, while the second focuses on multi-scale model architecture.
PETSA: Adapt Time-Series Forecasters at Test Time
PETSA is a parameter-efficient test-time adaptation method for time-series forecasting. It updates small input and output calibration modules rather than the full model, targeting non-stationary data with lower adaptation costs.
multiresolution-time-series-transformer: Forecast Time Series at Multiple Scales
A PyTorch forecasting model that processes time series at several temporal resolutions, then fuses the representations to predict future values. It suits experimentation with multi-scale forecasting, with adaptations from the paper that users should account for.
| PETSA | multiresolution-time-series-transformer | |
|---|---|---|
| Language | Shell | Python |
| License | NOASSERTION | — |
| Stars | 21 | 2 |
| Forks | 3 | 1 |
| Last analyzed | Oct 3, 2026 | Oct 3, 2026 |
Key differences
- PETSA is a test-time adaptation method for existing forecasting backbones, while multiresolution-time-series-transformer implements a multi-resolution forecasting model.
- PETSA updates low-rank input and output calibration modules with dynamic gating; multiresolution-time-series-transformer uses stride-based subsampling, separate transformer branches, and interpolation-based fusion.
- PETSA provides support for several named backbones and example scripts for listed datasets; multiresolution-time-series-transformer includes a training workflow that users can adapt to other dataset formats.
- PETSA is described as a Shell repository based on the Time-Series-Library, while multiresolution-time-series-transformer is a Python project using PyTorch.
- PETSA lists no asserted license, while multiresolution-time-series-transformer is described in its README as MIT-licensed, with a note to verify the repository license.
- The supplied metadata lists PETSA with 21 stars and 3 forks, and multiresolution-time-series-transformer with 2 stars and 1 fork.
Choose PETSA if you…
- evaluate parameter-efficient adaptation of pre-trained forecasters at inference time.
- compare adaptation across PETSA's supported backbones and listed datasets.
- investigate lower-cost alternatives to updating an entire forecasting model when data are non-stationary.
Choose multiresolution-time-series-transformer if you…
- prototype forecasting with high-, mid-, and low-resolution transformer branches.
- explore a compact PyTorch training and inference workflow for time-series forecasting.
- adapt the included training workflow to OHLCV or similar tabular time-series data.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.