[timeseries] Parallel Training of Time series Predictor

[shchur]

trafficstars

Discussed in https://github.com/autogluon/autogluon/discussions/4189

^{Originally posted by vamshik113 May 10, 2024} Hello, I've been utilizing AutoGluon and finding it quite enjoyable. I had a question about parallelizing model training. Is it possible to concurrently train all forecasters and then utilize the outputs of the best models for Ensemble Training? For instance, suppose I have an ECR Node group with multiple nodes that could be leveraged to parallelise the training of each model.

[Innixma]

Related: #4215

@rey-allan is currently looking into parallel training for Tabular with an identical goal. I think the solution we come up with should be transferable to TimeSeries with relative ease.

[jmakov]

Can't this be simply parallelized with ray.io?

[shchur]

@jmakov This will require several other modifications. We are not sure which path we will take and the specific timeline for this feature.

[jmakov]

Interesting, thought one would just refactor and add local ray cluster backend: https://github.com/autogluon/autogluon-cloud.