Torch 2.5.0 module 'torch.distributed' has no attribute 'init_process_group' in jetpack 6.1

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Orin
,
1

Issue:
I installed torch-2.5.0a0+872d972e41.nv24.8 and jetpack 6.1 in Agx orin,
found the torch lost module ‘torch.distributed’, the error is module ‘torch.distributed’ has no attribute ‘init_process_group’ as run following command in python

~$ python

Python 3.10.12 (main, Sep 11 2024, 15:47:36) [GCC 11.4.0] on linux
Type “help”, “copyright”, “credits” or “license” for more information.

import torch
print(torch.version)

2.5.0a0+872d972e41.nv24.08

import torch.distributed as dist
dist.init_process_group(backend=“nccl”)

Traceback (most recent call last):
File “”, line 1, in
AttributeError: module ‘torch.distributed’ has no attribute ‘init_process_group’

Environment:
device :Agx orin
torch verison : torch-2.5.0a0+872d972e41.nv24.08.17622132-cp310-cp310-linux_aarch64.whl download link selected from PyTorch for JetPack 6.1 as following link: Jetson Download Center | NVIDIA Developer
jatpack: 6.1
python:3.10.12

Question:
I tried to rebuild the pytorch 2.5.0 but failed, how to get full version of pytorch 2.5.0 using jetpack6.1 ?

4

Hi,

Please try the PyTorch package in the below link:

http://jetson.webredirect.org/jp6/cu126

We have confirmed that the package has been built with the distributed module.

Thanks.

5

Hi AastaLL:

I met a new issue as installed you mentioned new torch2.5.0, please refer to following error message:

import torch
import torch.distributed as dist
dist.init_process_group(backend=“nccl”)
Traceback (most recent call last):
File “”, line 1, in
File “/usr/local/lib/python3.10/dist-packages/torch/distributed/c10d_logger.py”, line 83, in wrapper
return func(*args, **kwargs)
File “/usr/local/lib/python3.10/dist-packages/torch/distributed/c10d_logger.py”, line 97, in wrapper
func_return = func(*args, **kwargs)
File “/usr/local/lib/python3.10/dist-packages/torch/distributed/distributed_c10d.py”, line 1520, in init_process_group
store, rank, world_size = next(rendezvous_iterator)
File “/usr/local/lib/python3.10/dist-packages/torch/distributed/rendezvous.py”, line 258, in _env_rendezvous_handler
rank = int(_get_env_or_raise(“RANK”))
File “/usr/local/lib/python3.10/dist-packages/torch/distributed/rendezvous.py”, line 243, in _get_env_or_raise
raise _env_error(env_var)
ValueError: Error initializing torch.distributed using env:// rendezvous: environment variable RANK expected, but not set

I tried in 3 environment installing new torch2.5.0, issues are same in Agx orin device , virtual environment and docker.

Thanks for you checking this issue !

6

Hi AastaLL:

Above issue should not be torch issue, the torch works very well as set os parameters as following.

import os
os.environ[‘MASTER_ADDR’] = ‘localhost’
os.environ[‘MASTER_PORT’] = ‘5678’
os.environ[‘RANK’] = ‘0’
os.environ[‘WORLD_SIZE’] = ‘1’
import torch
import torch.distributed as dist
dist.init_process_group(backend=“nccl”)

Thanks a lot for your help!!!

7

Hi,

Good to know it works now!
Thanks for the update.

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.