PatchTSMixer in HuggingFace

Hugging Face Blog · 2 years ago

IBM Research released PatchTSMixer, a lightweight time-series forecasting model based on MLP-Mixer architecture, in the Hugging Face Transformers library. The model outperforms state-of-the-art MLP and Transformer models by 8-60% in forecasting while using 2-3X less memory and runtime than Patch-Transformer models. Users can now apply PatchTSMixer to various downstream tasks including forecasting, classification, and regression through the Hugging Face implementation.