PaddleTS 1.0.2 Release Note EN

Release Highlights for PaddleTS 1.0.2

New Models

PaddleTS releases new time series representation model(s):

• Contrastive Learning of Disentangled Seasonal-trend Representations(CoST)

CPU & GPU Capable Docker Images for PaddleTS

• registry.baidubce.com/paddlets/paddlets:1.0.2
• registry.baidubce.com/paddlets/paddlets:1.0.2-gpu-cuda11.2-cudnn8

Pulling latest tag will get 1.0.2 release now.

• registry.baidubce.com/paddlets/paddlets:latest
• registry.baidubce.com/paddlets/paddlets:latest-gpu-cuda11.2-cudnn8

Changelog for PaddleTS 1.0.2

Legend for changelogs:

Feature: new features that PaddleTS does not support previously.

Optimization: an existing feature now either requires less computation or memory.

API Change: some changes that require you to change your code.

Fix: some features that previously did not work as stated in documents, or for reasonable expectations that should now work.

Changelog

paddlets.automl

• API Change autots.AutoTS::fit() removes old n_trails parameter and replace it with new n_trials parameter.
• Optimization automl.search_space_configer optimizes default search space for TCN deep time series forecasting model.

paddlets.models

• Feature PaddleTS introduces Contrastive Learning of Disentangled Seasonal-trend Representations(CoST) deep time series representation model.
• API Change models.forecasting.dl.nbeats.NBEATSModel replace num_blocks parameter type from old int with new Union[int, List[int]] to support more general use case. Thanks 2Bear for the contribution at #245.
• Feature PaddleTS now supports using newly released representation model CoST to solve time series forecasting task.

paddlets.metrics

• Fix Some internal issue for metrics is now fixed. #240 by YaoShuai.