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

Use CUDAPinnedPlace in buffered_reader #19112

Merged
merged 2 commits into from
Aug 14, 2019

Conversation

chengduoZH
Copy link
Contributor

@chengduoZH chengduoZH commented Aug 9, 2019

Make data transmission and model calculation overlap to make data transmission and model calculation overlap.

  1. The speed of fast_resnet model(multi-process 4x8 cards) has increased by about 5%.
  2. The speed of SE_ResNeXt50_32x4d model(single-process 8 cards) has increased by about 2.8%.
  3. The speed of TransformerBase model(single-process 8 cards) has increased by about 1.9%.
  4. The speed of TransformerBase model(single-process 1 cards) has increased by about 3..0%.

chengduozh added 2 commits August 9, 2019 21:16
test=develop
Copy link
Contributor

@gongweibao gongweibao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@chenwhql chenwhql self-requested a review August 14, 2019 07:40
Copy link
Contributor

@chenwhql chenwhql left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM for PADDLE_ENFORCE

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.

3 participants