Skip to content

cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 -> 8.4.1#215578

Merged
samuela merged 1 commit intoNixOS:masterfrom
ConnorBaker:cudnn-8.8.0
Feb 11, 2023
Merged

cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 -> 8.4.1#215578
samuela merged 1 commit intoNixOS:masterfrom
ConnorBaker:cudnn-8.8.0

Conversation

@ConnorBaker
Copy link
Contributor

@ConnorBaker ConnorBaker commented Feb 9, 2023

Description of changes

Adds the following versions:

  • 8.0.5
  • 8.2.4
  • 8.7.0

Bumps the following versions:

  • 8.3.2 -> 8.3.3
  • 8.4.0 -> 8.4.1

Corrects supported CUDA versions:

  • 8.3.x does not support CUDA 11.7 (stops at 11.6)
  • 8.1.1 supports CUDA 11.0 and 11.1 (in addition to 11.2)

Changes default CUDNN versions:

  • CUDA 10.1 now uses the last supported release, 8.0.5
  • CUDA 10.2 now uses the last supported release, 8.7.0
  • CUDA 11.x now use 8.7.0
Things done
  • Built on platform(s)
    • x86_64-linux
    • aarch64-linux
    • x86_64-darwin
    • aarch64-darwin
  • For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
  • Tested, as applicable:
  • Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
  • Tested basic functionality of all binary files (usually in ./result/bin/)
  • 23.05 Release Notes (or backporting 22.11 Release notes)
    • (Package updates) Added a release notes entry if the change is major or breaking
    • (Module updates) Added a release notes entry if the change is significant
    • (Module addition) Added a release notes entry if adding a new NixOS module
    • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
  • Fits CONTRIBUTING.md.

@ConnorBaker ConnorBaker mentioned this pull request Feb 9, 2023
13 tasks
@samuela
Copy link
Member

samuela commented Feb 9, 2023

LGTM is there a reason we need to wait on 8.8.0? I feel like we could also create another PR for that once it becomes available

@ConnorBaker
Copy link
Contributor Author

Fair enough! I'll remove it and clean this up.

@ConnorBaker ConnorBaker changed the title [DRAFT] cudnn: init 8.0.5, 8.2.4, 8.7.0, 8.8.0; update 8.3.2 -> 8.3.3, 8.4.0 … [DRAFT] cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 … Feb 9, 2023
@ConnorBaker ConnorBaker changed the title [DRAFT] cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 … [DRAFT] cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 -> 8.4.1 Feb 9, 2023
@ConnorBaker ConnorBaker changed the title [DRAFT] cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 -> 8.4.1 cudnn: init 8.0.5, 8.2.4, 8.7.0; update 8.3.2 -> 8.3.3, 8.4.0 -> 8.4.1 Feb 9, 2023
@ConnorBaker ConnorBaker marked this pull request as ready for review February 9, 2023 23:14
@ConnorBaker
Copy link
Contributor Author

@samuela who should I tag for review in general on CUDA-related issues? The CUDA maintainers team listed at the beginning of https://nixos.wiki/wiki/CUDA gives a 404: https://github.com/orgs/NixOS/teams/cuda-maintainers. Does the tag still work? When I tried typing it, GitHub didn't provide an auto-complete so I wasn't sure if it's valid.

@ofborg ofborg bot added 8.has: clean-up This PR removes packages or removes other cruft 8.has: package (new) This PR adds a new package 10.rebuild-darwin: 1-10 This PR causes between 1 and 10 packages to rebuild on Darwin. 10.rebuild-linux: 11-100 This PR causes between 11 and 100 packages to rebuild on Linux. labels Feb 9, 2023
@nixos-discourse
Copy link

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/how-to-install-a-specific-version-of-cuda-and-cudnn/21725/6

@SomeoneSerge
Copy link
Contributor

@NixOS/cuda-maintainers is the right tag, maybe it's something about the github organization settings?

@SomeoneSerge SomeoneSerge added the 6.topic: cuda Parallel computing platform and API label Feb 10, 2023
@samuela
Copy link
Member

samuela commented Feb 11, 2023

Result of nixpkgs-review pr 215578 run on x86_64-linux 1

1 package marked as broken and skipped:
  • python310Packages.caffeWithCuda
4 packages failed to build:
  • caffeWithCuda
  • cudaPackages.tensorrt (cudaPackages.tensorrt_8_4_0)
  • python310Packages.tensorrt
  • python311Packages.tensorrt
11 packages built:
  • cudaPackages.cudnn (cudaPackages.cudnn_8_7_0)
  • cudaPackages.cudnn_8_4_1
  • katagoWithCuda
  • python310Packages.TheanoWithCuda
  • python310Packages.cupy
  • python310Packages.jaxlibWithCuda
  • python310Packages.tensorflowWithCuda
  • python310Packages.torchWithCuda
  • python311Packages.TheanoWithCuda
  • python311Packages.cupy
  • python311Packages.jaxlibWithCuda

@ConnorBaker
Copy link
Contributor Author

Woah nixpkgs-review is neat! I'm running it to try to reproduce the failures.

First thing I've seen so far is this, but it's still going.

[connorbaker@fedora nixpkgs]$ nixpkgs-review pr 215578
$ git -c fetch.prune=false fetch --no-tags --force https://github.com/NixOS/nixpkgs master:refs/nixpkgs-review/0 pull/215578/head:refs/nixpkgs-review/1
remote: Enumerating objects: 1824, done.
remote: Counting objects: 100% (1119/1119), done.
remote: Compressing objects: 100% (86/86), done.
remote: Total 1824 (delta 1060), reused 1067 (delta 1028), pack-reused 705
Receiving objects: 100% (1824/1824), 1.36 MiB | 46.56 MiB/s, done.
Resolving deltas: 100% (1285/1285), completed with 302 local objects.
From https://github.com/NixOS/nixpkgs
 * [new branch]              master                -> refs/nixpkgs-review/0
 * [new ref]                 refs/pull/215578/head -> refs/nixpkgs-review/1
$ git worktree add /home/connorbaker/.cache/nixpkgs-review/pr-215578/nixpkgs 592d3352e74deaa2a97fb31db36163d30a92eb36
Preparing worktree (detached HEAD 592d3352e74)
HEAD is now at 592d3352e74 Merge pull request #215823 from fabaff/renault-api-bump
$ git merge --no-commit --no-ff 56fd717ee3069dc07805afc78ca6af5f02e4ac9b
Automatic merge went well; stopped before committing as requested
$ nix --extra-experimental-features nix-command no-url-literals build --no-link --keep-going --no-allow-import-from-derivation --option build-use-sandbox relaxed -f /home/connorbaker/.cache/nixpkgs-review/pr-215578/build.nix
error: builder for '/nix/store/ixb1rldrrb92848s2sgvkd5kx3p3xgxn-TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz.drv' failed with exit code 1;
       last 10 log lines:
       > download the 8.4.0.6 Linux x86_64 TAR package for CUDA 11.7 from
       > https://developer.nvidia.com/tensorrt.
       >
       > Once you have downloaded the file, add it to the store with the following
       > command, and try building this derivation again.
       >
       > $ nix-store --add-fixed sha256 TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz
       >
       > ***
       >
       For full logs, run 'nix log /nix/store/ixb1rldrrb92848s2sgvkd5kx3p3xgxn-TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz.drv'.
[4/11/47 built (1 failed), 207 copied (3856.2/3857.0 MiB), 1394.5 MiB DL] building cuda_10.1.243_418.87.00_linux.run: 100 2453M  100 2453M    0     0  39.7M      0  0:01:01  0:01:01 --:--:-- 84.5M

@samuela
Copy link
Member

samuela commented Feb 11, 2023

Yeah, nixpkgs-review is a life saver!

TensorRT always fails for that same reason in nixpkgs-review, and caffeWithCuda is marked as broken anyhow, so this looks like a success to me.

Thanks for contributing @ConnorBaker!

@samuela samuela merged commit 441c0e5 into NixOS:master Feb 11, 2023
@ConnorBaker ConnorBaker deleted the cudnn-8.8.0 branch February 11, 2023 19:07
@ConnorBaker ConnorBaker mentioned this pull request Feb 20, 2023
12 tasks
@nixos-discourse
Copy link

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/tweag-nix-dev-update-45/26397/1

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

6.topic: cuda Parallel computing platform and API 8.has: clean-up This PR removes packages or removes other cruft 8.has: package (new) This PR adds a new package 10.rebuild-darwin: 1-10 This PR causes between 1 and 10 packages to rebuild on Darwin. 10.rebuild-linux: 11-100 This PR causes between 11 and 100 packages to rebuild on Linux.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants