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

python312Packages.torch[-bin]: 2.5.1 -> 2.6.0 #377785

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

Conversation

GaetanLepage
Copy link
Contributor

@GaetanLepage GaetanLepage commented Jan 29, 2025

Things done

Update the torch ecosystem:

x86_64-linux aarch64-linux x86_64-darwin aarch64-darwin
torch
torch (CUDA)
torch (ROCm) ✝️ ✝️ ✖️ ✖️
torch-bin ✖️
torchaudio
torchaudio (CUDA)
torchaudio (ROCm)
torchaudio-bin ✖️
torchvision
torchvision (CUDA)
torchvision (ROCm)
torchvision-bin ✖️
  • ✅: building
  • ❌: failing
  • ✝️: marked as broken
  • ✖️: not available

  • Built on platform(s)
    • x86_64-linux
    • aarch64-linux
    • x86_64-darwin
    • aarch64-darwin
  • For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
    • sandbox = relaxed
    • sandbox = true
  • 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/)
  • 25.05 Release Notes (or backporting 24.11 and 25.05 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
  • Fits CONTRIBUTING.md.

closes #378424


Add a 👍 reaction to pull requests you find important.

@github-actions github-actions bot added 6.topic: python 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin 10.rebuild-linux: 0 This PR does not cause any packages to rebuild on Linux labels Jan 29, 2025
@GaetanLepage
Copy link
Contributor Author

GaetanLepage commented Jan 29, 2025

torch-bin failing with:

auto-patchelf: 1 dependencies could not be satisfied
error: auto-patchelf could not satisfy dependency libcusparseLt.so.0 wanted by /nix/store/1lr2jw9kl3p1kkjshz7ihs8mwgvw556i-python3.12-torch-2.6.0/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so
auto-patchelf failed to find all the required dependencies.

cc @SomeoneSerge @ConnorBaker do you know which library needs to be added ?

@github-actions github-actions bot added 10.rebuild-darwin: 1-10 10.rebuild-linux: 1-10 and removed 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin 10.rebuild-linux: 0 This PR does not cause any packages to rebuild on Linux labels Jan 29, 2025
@GaetanLepage
Copy link
Contributor Author

@zeuner I'm having some issues with the process you have introduced in #277789.
Running the unroll-src script fails with:

❮ ./result 2.6.0
{
  version,
  fetchFromGitLab,
  fetchFromGitHub,
  runCommand,
}:
assert version == "2.6.0";
Loading required package: processx
Loading required package: jsonlite
Loading required package: stringr
Loading required package: plyr
Loading required package: dplyr

Attaching package: ‘dplyr’

The following objects are masked from ‘package:plyr’:

    arrange, count, desc, failwith, id, mutate, rename, summarise,
    summarize

The following objects are masked from ‘package:stats’:

    filter, lag

The following objects are masked from ‘package:base’:

    intersect, setdiff, setequal, union

Error in `str_match()`:
! `pattern` must be a regular expression.
Backtrace:
    ▆
 1. └─global add_missing_submodules()
 2.   └─global repo_submodules(fetched)
 3.     └─stringr::str_match(trimmed, "")
 4.       └─cli::cli_abort("`pattern` must be a regular expression.")
 5.         └─rlang::abort(...)
Execution halted

# Update using: unroll-src [version]

@zeuner
Copy link
Contributor

zeuner commented Jan 30, 2025

@zeuner I'm having some issues with the process you have introduced in #277789. Running the unroll-src script fails with:

❮ ./result 2.6.0
{
  version,
  fetchFromGitLab,
  fetchFromGitHub,
  runCommand,
}:
assert version == "2.6.0";
Loading required package: processx
Loading required package: jsonlite
Loading required package: stringr
Loading required package: plyr
Loading required package: dplyr

Attaching package: ‘dplyr’

The following objects are masked from ‘package:plyr’:

    arrange, count, desc, failwith, id, mutate, rename, summarise,
    summarize

The following objects are masked from ‘package:stats’:

    filter, lag

The following objects are masked from ‘package:base’:

    intersect, setdiff, setequal, union

Error in `str_match()`:
! `pattern` must be a regular expression.
Backtrace:
    ▆
 1. └─global add_missing_submodules()
 2.   └─global repo_submodules(fetched)
 3.     └─stringr::str_match(trimmed, "")
 4.       └─cli::cli_abort("`pattern` must be a regular expression.")
 5.         └─rlang::abort(...)
Execution halted

# Update using: unroll-src [version]

I think it will work using the latest git-unroll, updated in #377956

@GaetanLepage
Copy link
Contributor Author

Fails on aarch64-linux:

/build/pytorch/third_party/XNNPACK/src/f16-avgpool/f16-avgpool-9p8x-minmax-neonfp16arith-c8.c:31:64: error: passing argument 1 of ‘vld1q_dup_u16’ from incompatible pointer type [-Wincompatible-pointer-types]
   31 |   const float16x8_t vmin = vreinterpretq_f16_u16(vld1q_dup_u16(&params->scalar.min));
      |                                                                ^~~~~~~~~~~~~~~~~~~
      |                                                                |
      |                                                                const xnn_float16 * {aka const _Float16 *}

GCC 14 ?

@nix-owners nix-owners bot requested a review from junjihashimoto January 30, 2025 13:33
@GaetanLepage GaetanLepage marked this pull request as ready for review January 30, 2025 13:43
@GaetanLepage
Copy link
Contributor Author

GaetanLepage commented Jan 30, 2025

It looks like libcusparseLt.so is now a hard dependency:

error: auto-patchelf could not satisfy dependency libcusparseLt.so.0 wanted by /nix/store/1lr2jw9kl3p1kkjshz7ihs8mwgvw556i-python3.12-torch-2.6.0/lib/python3.12/site-packages/torch/lib/libtorch_cuda.so

If I understand correctly, we need to wait for #306172.
See #270446.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants