avalanche: Examples memory error on Windows

I tested the examples on my Windows machines. Most of them work fine, but some of them throw a memory error (e.g. examples\joint_training.py):

raceback (most recent call last):
  File "<string>", line 1, in <module>
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\multiprocessing\spawn.py", line 116, in spawn_main
    exitcode = _main(fd, parent_sentinel)
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\multiprocessing\spawn.py", line 125, in _main
    prepare(preparation_data)
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\multiprocessing\spawn.py", line 236, in prepare
    _fixup_main_from_path(data['init_main_from_path'])
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\multiprocessing\spawn.py", line 287, in _fixup_main_from_path
    main_content = runpy.run_path(main_path,
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\runpy.py", line 265, in run_path
    return _run_module_code(code, init_globals, run_name,
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\runpy.py", line 97, in _run_module_code
    _run_code(code, mod_globals, init_globals,
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "C:\Users\w-32\OneDrive - University of Pisa\Uni\avalanche\examples\joint_training.py", line 26, in <module>
    import torch
  File "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\site-packages\torch\__init__.py", line 116, in <module>
    raise err
OSError: [WinError 1455] Il file di paging è troppo piccolo per essere completato. Error loading "C:\Users\w-32\Anaconda3\envs\avalanche-env\lib\site-packages\torch\lib\caffe2_detectron_ops_gpu.dll" or one of its dependencies.

I have 16GB of RAM, and looking at the task manager it seems that there is still memory available.

On the remote server (linux machine, 255GB RAM) everything works.

About this issue

  • Original URL
  • State: closed
  • Created 3 years ago
  • Comments: 16 (7 by maintainers)

Most upvoted comments

I tested the example on a different Windows machine and it seems to work. Maybe this is a problem with my machine and not avalanche, but it’s a really weird memory problem

I may have found the problem. Apparently, Windows does not work well with num_workers > 0. This is a known issue (see here). Can we set num_workers=0 in the examples? It shouldn’t have a big impact for the small datasets used in the examples.

@Ayshine Sorry, wrong command, it should be python -m examples.joint_training