Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Time-series experiments #4

Closed
AndreRoehrig opened this issue Dec 19, 2021 · 2 comments
Closed

Time-series experiments #4

AndreRoehrig opened this issue Dec 19, 2021 · 2 comments

Comments

@AndreRoehrig
Copy link

The experiments using datasets like etth, ettm, and ecl don't run (despite the definitions in /config). I think it's because the datasets cannot be downloaded automatically like the LRA experiments.

Would it be possible to add this, or explain how these experimets could be run?

@AndreRoehrig
Copy link
Author

python -m train wandb=null experiment=s4-informer-etth
dataset appears more than once in the final defaults list

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.

@krandiash
Copy link
Contributor

Hi, apologies about the delay, we've added the relevant code required to rerun these experiments. Please try the following e.g. to run ECL 168 length univariate:

python -m train experiment=s4-informer-ecl decoder.mode=last dataset.size=[24,24,168] dataset.timeenc=1 dataset.variant=0 dataset.features=S model.n_layers=2 model.dropout=0.25 optimizer.lr=0.01 task.metrics=[mse,mae]

Data for the experiments can be found here. Make a data folder inside the state-spaces repo, and unzip the informer.zip file from the link inside that folder.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants