Ubuntu22.04如何升级？

AKA

该容器用的是Ubuntu20.04，我在部署OpenSora项目时出现很多兼容依赖问题，我想知道如何将其升级为22.04版本（我自己尝试过但经常升级过程中报错），我想知道有解决办法吗？或者平台内有虚拟机22.04版本，但没有相关的文档，我该怎样配置成图中类似的环境和界面？

yorda

AKA 可以找管理员安装。方便的话请告知一下是哪个平台

xwhuan

AKA 遇到的兼容依赖问题也可以截图发出来

AKA

yorda 河南大学高性能计算平台，麻烦安一下吧谢谢。https://zsglxt.henu.edu.cn/user/instance

xwhuan

AKA 可以提供下OpenSora项目的安装链接和环境需求，我们来安装

AKA

xwhuan https://github.com/hpcaitech/Open-Sora这是项目链接

AKA

create a virtual env and activate (conda as an example)

conda create -n opensora python=3.10
conda activate opensora

download the repo

git clone https://github.com/hpcaitech/Open-Sora
cd Open-Sora

Ensure torch >= 2.4.0

pip install -v . # for development mode, pip install -v -e .
pip install xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121 # install xformers according to your cuda version
pip install flash-attn --no-build-isolation

这是项目安装的代码，要求只有torch >= 2.4.0和python=3.10，但这都可以自己搭建，需求Ubuntu22.04是下面的flash-attn导致的，他下载的时候是编译好的版本，他编译版本是在Ubuntu22.04（大概），导致编译版本在Ubuntu20.04.无法读取，报错一直说glibc一直版本过低，想绕开自己编译的时候又一直编译不了。
总结目前需求是Ubuntu22.04 cuda版本12以上

https://github.com/hpcaitech/Open-Sora这是项目安装链接

xwhuan

AKA 好的，我看下怎么处理

xwhuan

AKA
可以试下conda安装flash-attn，可以安装成功
pip install -v .
pip install xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu121

conda install conda-forge::flash-attn

AKA

xwhuan 这个下载会改变torch环境为2.8导致xformers不兼容，然后我又去官网找了对应的2.8的xformers，下载完成后在执行demo示例的时候出现了torch和torchvision不兼容的情况，具体在上面的楼里写的有，感觉真得装一个Ubuntu22.04，不然这不停地下载依赖不停地出现兼容问题，搞不完。

AKA

(opensora) studentluo@n1:~/jupyterlab/Open-Sora$ conda install conda-forge::flash-attn
Channels:

defaults
conda-forge
Platform: linux-64
Collecting package metadata (repodata.json): done
Solving environment: done

Package Plan

environment location: /home/studentluo/.conda/envs/opensora

added / updated specs:
- conda-forge::flash-attn

The following NEW packages will be INSTALLED:

flash-attn conda-forge/linux-64::flash-attn-2.8.3-py310h295a8a3_1
pytorch conda-forge/linux-64::pytorch-2.8.0-cuda129_mkl_py310_h4bbb3bc_302
sympy pkgs/main/linux-64::sympy-1.14.0-py310h06a4308_0

Proceed ([y]/n)? 这个指令会导致torch环境变为2.8，导致其他依赖不兼容。如果可以的话我还是觉得直接配一个Ubuntu22.04的容器环境，然后按文档使用pip指令会好很多。我之前在别的平台用4090部署成功过，最后只有显存不够的问题。

xwhuan

AKA 容器22.04目前还在开发适配，还需要一定的时间才能上线平台，先试试虚拟机ubuntu22吧

AKA

(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora$ python -c "import torch; print(torch.version)" # 应输出 2.4.0

2.4.0+cu121
(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora$ python -c "import torchvision; print(torchvision.version)" # 应输出 0.19.0
0.19.0+cu121
(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora$ python -c "import xformers; print('xformers OK')"
xformers OK这是我按照文档在执行conda install conda-forge::flash-attn前的环境

AKA

这是下载flash-attn之后的环境
(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora$ python -c "import torch; print(torch.version)"

2.8.0
(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora python -c "import xformers; print('xformers OK')"
WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for: PyTorch 2.4.0+cu121 with CUDA 1201 (you have 2.8.0) Python 3.10.14 (you have 3.10.19) Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers) Memory-efficient attention, SwiGLU, sparse and more won't be available. Set XFORMERS_MORE_DETAILS=1 for more details xformers OK

(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora python -c "import flash_attn; print('flash OK')"
flash OK

AKA

总之我倾向于部署一个Ubuntu22.04用pip指令会轻松很多，不然不断的找配套环境下载过程中会不断出现依赖问题。

我在查阅了官方文档（https://github.com/facebookresearch/xformers/releases）后，发现v0.0.32: Wheels for PyTorch 2.8.0，
我现在使用这个指令pip install xformers==0.0.32.post2 --no-deps下载了符合torch2.8环境的xformers。
后续模型使用过程中又出现了新问题
核心错误信息：RuntimeError: operator torchvision::nms does not exist
RuntimeError: 程序成功启动，但在运行过程中遇到了一个致命错误。
operator ... does not exist: 一个函数（操作符）丢失了。
torchvision::nms: 丢失的函数是 torchvision 库里的 nms（非极大值抑制）。这是一个核心的图像处理函数。
分析大概率是PyTorch 和 Torchvision 之间不匹配（因为我们前面用conda下载flash-attn的时候会改变torch环境）
以下是代码部分：
(opensora-conda-forge) studentluo@n1:~/jupyterlab/Open-Sora$ torchrun --nproc_per_node 2 --standalone scripts/diffusion/inference.py configs/diffusion/inference/256px.py \

--ckpt-path YOUR_256px_MODEL_PATH \
--cond_type i2v_head \
--prompt "A plump pig wallows in a muddy pond on a rustic farm, its pink snout poking out as it snorts contentedly. The camera captures the pig's playful splashes, sending ripples through the water under the midday sun. Wooden fences and a red barn stand in the background, framed by rolling green hills. The pig's muddy coat glistens in the sunlight, showcasing the simple pleasures of its carefree life." \
--ref assets/texts/i2v.png
W1114 14:26:30.211000 10317 site-packages/torch/distributed/run.py:774]
W1114 14:26:30.211000 10317 site-packages/torch/distributed/run.py:774] *****************************************
W1114 14:26:30.211000 10317 site-packages/torch/distributed/run.py:774] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
W1114 14:26:30.211000 10317 site-packages/torch/distributed/run.py:774] *****************************************
Traceback (most recent call last):
File "/home/studentluo/jupyterlab/Open-Sora/scripts/diffusion/inference.py", line 15, in <module>
from opensora.datasets.dataloader import prepare_dataloader
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/opensora/datasets/init.py", line 1, in <module>
from .datasets import TextDataset, VideoTextDataset
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/opensora/datasets/datasets.py", line 8, in <module>
from torchvision.datasets.folder import pil_loader
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torchvision/init.py", line 10, in <module>
Traceback (most recent call last):
File "/home/studentluo/jupyterlab/Open-Sora/scripts/diffusion/inference.py", line 15, in <module>
from torchvision import meta_registrations, datasets, io, models, ops, transforms, utils # usort:skip
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torchvision/meta_registrations.py", line 164, in <module>
from opensora.datasets.dataloader import prepare_dataloader
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/opensora/datasets/init.py", line 1, in <module>
def meta_nms(dets, scores, iou_threshold):
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library.py", line 1069, in register
from .datasets import TextDataset, VideoTextDataset
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/opensora/datasets/datasets.py", line 8, in <module>
from torchvision.datasets.folder import pil_loader
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torchvision/init.py", line 10, in <module>
use_lib._register_fake(
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library.py", line 219, in _register_fake
from torchvision import meta_registrations, datasets, io, models, ops, transforms, utils # usort:skip
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torchvision/meta_registrations.py", line 164, in <module>
handle = entry.fake_impl.register(
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library/fake_impl.py", line 50, in register
def meta_nms(dets, scores, iou_threshold):
if torch.C._dispatch_has_kernel_for_dispatch_key(self.qualname, "Meta"): File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library.py", line 1069, in register

RuntimeError: operator torchvision::nms does not exist
use_lib._register_fake(
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library.py", line 219, in register_fake
handle = entry.fake_impl.register(
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/library/fake_impl.py", line 50, in register
if torch.C.dispatch_has_kernel_for_dispatch_key(self.qualname, "Meta"):
RuntimeError: operator torchvision::nms does not exist
E1114 14:26:32.970000 10317 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 10321) of binary: /home/studentluo/.conda/envs/opensora-conda-forge/bin/python
Traceback (most recent call last):
File "/home/studentluo/.conda/envs/opensora-conda-forge/bin/torchrun", line 9, in <module>
sys.exit(main())
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/init.py", line 357, in wrapper
return f(*args, **kwargs)
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/distributed/run.py", line 901, in main
run(args)
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/distributed/run.py", line 892, in run
elastic_launch(
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 143, in call
return launch_agent(self.config, self.entrypoint, list(args))
File "/home/studentluo/.conda/envs/opensora-conda-forge/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 277, in launch_agent
raise ChildFailedError(

torch.distributed.elastic.multiprocessing.errors.ChildFailedError:

scripts/diffusion/inference.py FAILED

Failures:
[1]:
time : 2025-11-14_14:26:32
host : n1.example.com
rank : 1 (local_rank: 1)
exitcode : 1 (pid: 10322)
error_file: <N/A>

traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html

Root Cause (first observed failure):
[0]:
time : 2025-11-14_14:26:32
host : n1.example.com
rank : 0 (local_rank: 0)
exitcode : 1 (pid: 10321)
error_file: <N/A>

traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html

AKA

xwhuan 请问大概需要多久时间？一个星期or一个月？

xwhuan

AKA 这个开发周期还挺长的，近期应该没法上线。我再看看这个项目具体的软件依赖版本需求吧