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.

Sign up for GitHub

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

Update tsfresh_on_a_cluster.rst #1069

Merged
merged 1 commit into from
May 26, 2024
Merged

Update tsfresh_on_a_cluster.rst #1069

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions docs/text/tsfresh_on_a_cluster.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -229,3 +229,11 @@ This issue can be solved by constraining the C libraries to single threads, usin
Put these lines at the beginning of your notebook/python script - before you call any tsfresh code or import any other module.

The more cores your host computer has, the more improvement in processing speed will be gained by implementing these environment changes. Speed increases of between 6x and 26x have been observed depending on the type of the host machine.

.. NOTE::
If you intend to run machine learning pipelines after feature extraction, it is highly recommended to revert these changes.
Settings have been tested on GPU-accelerated XGBoost classifier training task with 9x speed reduction.
Many popular machine learning libraries like scikit-learn and Tensorflow/PyTorch rely on parallelization of your CPU cores to speed up CPU-dependent low-level computations.
By forcing those libraries to only use a single thread, it creates a bottleneck for other downstream tasks even if you are accelerating them with GPUs.
e.g. CPU-computation of loss function after every PyTorch neural network forward pass will only run on one thread, while the GPU idles and waits for its completion before starting on its backpropogation tasks.