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

{lib}[fosscuda/2020b] TensorRT v8.0.1.6 w/ Python 3.8.6 #13499

Conversation

robert-mijakovic
Copy link
Contributor

@robert-mijakovic robert-mijakovic commented Jul 24, 2021

(created using eb --new-pr)

@robert-mijakovic robert-mijakovic changed the title {lib}[fosscuda/2021a] TensorRT v8.0.1.6 w/ Python 3.9.5 {lib}[fosscuda/2020b] TensorRT v8.0.1.6 w/ Python 3.8.6 Jul 27, 2021
@lexming lexming added the update label Jul 28, 2021
@lexming lexming added this to the 4.x milestone Jul 28, 2021
@lexming
Copy link
Contributor

lexming commented Jul 28, 2021

Test report by @lexming
SUCCESS
Build succeeded for 2 out of 2 (1 easyconfigs in total)
node226.hydra.os - Linux centos linux 7.9.2009, x86_64, Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz, Python 2.7.5
See https://gist.github.com/20ca5d7300c9ac1a3eccc336a00acf42 for a full test report.

lexming
lexming previously approved these changes Jul 28, 2021
Copy link
Contributor

@lexming lexming left a comment

Choose a reason for hiding this comment

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

Even though the name of the sources say cuda-11.3, this tarball for x86_64 is compatible with CUDA v11.1 as well. This is not the case with the tarballs for ARM and Power though, those are only compatible with CUDA v11.3.

LGTM

@lexming lexming modified the milestones: 4.x, next release (4.4.2?) Jul 28, 2021
Copy link
Contributor

@lexming lexming left a comment

Choose a reason for hiding this comment

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

After further inspection, it seems that there is a soft requirement of TensorRT 8.0 on cuDNN v8.2 and we might be going into uncharted territory by using cuDNN v8.0.4. So I think that we should downgrade to TensorRT 7.2.2, which more closely matches the available dependencies in fosscuda/2020b, namely Python 3.8 and cuDNN 8.0.

If you agree with this change, I made a PR with the downgrade in robert-mijakovic#2

@lexming lexming dismissed their stale review July 29, 2021 21:37

new info provided

@boegel boegel modified the milestones: 4.4.2, release after 4.4.2 Sep 3, 2021
@boegel boegel modified the milestones: 4.5.0 (next release), 4.x Oct 28, 2021
downgrade TensorRT in fosscuda/2020b to version 7.2.2.3
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants