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

Added performance flags for better inference performance on Xeon #760

Open
wants to merge 3 commits into
base: master
Choose a base branch
from

Conversation

vivek-rane
Copy link

I'm a part of the TensorFlow optimization team at Intel and we're working on improving performance of YoloV2/darkflow on Xeon. As a first step, I'm adding the ability to measure and tweak performance. We will also contribute changes to TensorFlow and MKL-DNN to speed up YoloV2/DarkFlow.

  • Added flags for getting better performance during inference when running on TensorFlow built with MKL-DNN
  • Added a script to generate reasonable defaults for these flags depending on the system being used
  • Added ability to generate timeline to find performance bottlenecks

With the right settings, inference performance on a 28 core Xeon Scalable Processor goes up by more than 60%

@vivek-rane
Copy link
Author

vivek-rane commented May 15, 2018

Hang on - I tested this with python 2.7�. Let me update the code for 3.5.
Edit: done

@vivek-rane
Copy link
Author

Could someone help me with the failure in the CI? If I change the offset in darkflow/utils/loader.py from 16 to 20, it works. Not sure why.

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

Successfully merging this pull request may close these issues.

1 participant