search

QData / spacetimeformer

Multivariate Time Series Forecasting with efficient Transformers. Code for the paper "Long-Range Transformers for Dynamic Spatiotemporal Forecasting."
https://arxiv.org/abs/2109.12218
MIT License
808 stars 191 forks source link

Training times. #14

Closed InterHato closed 2 years ago

InterHato commented 2 years ago

For the example commands stated in the repository what are the training times that we should be expecting. And what were the hardware specification these tests were run on, with 4 gpu's specified keeping in mind.

jakegrigsby commented 2 years ago

Sorry for the late reply, I missed this one.

Our lab servers have a mix of different GPUs, but I looked through some of the Metr-LA log files and it looks like on 4 NVIDIA GeForce GTX 1080 Ti (11 GB) a training run takes a little over three hours.