-
Notifications
You must be signed in to change notification settings - Fork 31
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
Add workflow example for time series forecasting #255
Conversation
hcho3
commented
Jul 27, 2023
- Use data from the M5 competition, a well-known competition for time series forecasting
- Use Dask Kubernetes to run hyperparameter optimization
- Use GPU-accelerated XGBoost and cuDF to accelerate time series analysis end-to-end
- All preprocessing logic uses cuDF.
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
source/examples/time-series-forecasting-with-hpo/preprocessing_part1.ipynb
Outdated
Show resolved
Hide resolved
source/examples/time-series-forecasting-with-hpo/preprocessing_part2.ipynb
Outdated
Show resolved
Hide resolved
@jacobtomlinson For the time being, I put in the extra notebooks under the |
source/examples/time-series-forecasting-with-hpo/preprocessing_part1.ipynb
Outdated
Show resolved
Hide resolved
source/examples/time-series-forecasting-with-hpo/preprocessing_part1.ipynb
Outdated
Show resolved
Hide resolved
source/examples/time-series-forecasting-with-hpo/start_here.ipynb
Outdated
Show resolved
Hide resolved
source/examples/time-series-forecasting-with-hpo/training_and_evaluation.ipynb
Outdated
Show resolved
Hide resolved
source/examples/time-series-forecasting-with-hpo/training_and_evaluation.ipynb
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @hcho3 sorry it has taken me so long to get a review on this one.
Overall I think the structure is great, I've added a few comments.
One high-level question I had is why are there so many separate pre-processing notebooks? Some of them seem to load the data, do some small transformation and save them again. I think it would be more pleasant to read if there were fewer (or just one) pre-processing notebook. What do you think?
Also can you add some tags to the start-here
notebook?
source/examples/time-series-forecasting-with-hpo/start_here.ipynb
Outdated
Show resolved
Hide resolved
8319a78
to
4ee82d9
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is looking really great. Thanks for taking the time to merge this into one notebook and
address the review feedback.
The only small nit I have is that the DISABLE_JUPYTER
env var is no longer needed, so you can remove that from the code cell.