Skip to main content

Build fast data processing pipelines easily

Project description


Many tasks in machine learning, deep learning and other fields require complex data processing that takes a lot of time. Ideally, this processing should run in parallel to the main process, preparing data for usage (by neural net, for instance). PDP provide simple interface to organize pipeline of data processing with simple blocks that satisfy most typical needs.

Use cases

  • Neural Net training, where you need a way to train net, load data from the disk and augment it. PDP allows user to do all these things at the same time without need to use threading module directly.


Are in repository in examples folder

Is it fast?

Speed and parallel execution is a top priority. Right now threads are used to exchange information between pipline stages, because it’s memory and CPU efficient to exchange data between threads and not processes. Python’s threads are flawed by GIL, but it doesn’t affect performance for IO-bound tasks and for numpy operations. Since all operations for data augmentations are likely to be done in numpy operations, performance will not be significantly affected by GIL.


pip install pdp

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pdp-0.3.0.tar.gz (5.2 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page