mkl<=2024.0 CI pin installs a libtorch build that won't import on glibc 2.41+

未关闭
#2,539 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
2/5
预计耗时
1-3 小时
新手友好度
54/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
github-actions, python, ubuntu

调研方向

Start with .github/workflows/test.yml and inspect the pytorch link job's package pin and runner configuration; the issue cites conda-forge/pytorch-cpu-feedstock#350 and #355 for the fixed builds. Confirm which mkl constraint allows a fixed libtorch build to install on the target Ubuntu runner, then check that import torch succeeds in the job. #2537 pins the runner to Ubuntu 24.04 as a temporary workaround.

由索引模型根据 Issue 内容生成。

描述

GitHub CI/CD torch

import torch fails in the pytorch link CI job since ubuntu-latest moved to Ubuntu 26.04, with ImportError: libtorch_cpu.so: cannot enable executable stack. The "mkl<=2024.0" pin in .github/workflows/test.yml limits conda-forge's libtorch 2.5.1 to build _107, which predates the executable-stack fix for conda-forge/pytorch-cpu-feedstock#350 (conda-forge/pytorch-cpu-feedstock#355). Every later CPU build of 2.5.1 requires mkl >=2024.2.2, so the job can't get a fixed build while the pin stays.

The last passing run was on ubuntu-24.04 at 2026-10-19 18:15 UTC. Every run since 2026-10-20 01:14 UTC has landed on ubuntu-26.04 and failed, with the same _107 build installed in both. Example: https://github.com/pymc-devs/pytensor/actions/runs/39483309345/job/118837516328. #2537 pins the job to ubuntu-24.04 in the meantime.

主要语言
Python
星标
654
派生
258
平均合并
3 天 41 分钟
30 天内合并 PR
42

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

pymc-devs/pytensor 的其他 Issue

查看 pymc-devs/pytensor 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。