You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As said in the title, training stops at exception.
In mmengine/fileio/backends/local_backend.py, line 416, I've changed into following to temp fix it.
try:
ifnotself.exists(filepath):
raiseFileNotFoundError(f'filepath {filepath} does not exist')
ifself.isdir(filepath):
raiseIsADirectoryError('filepath should be a file')
except:
print("not exist!!!", filepath)
try:
os.remove(filepath)
except:
print("already removed!!!", filepath)
Hi @serser, you can run the command, mim install mmengine -U or pip install mmengine -U, to install the latest mmengine (0.3.1) which will solve this problem.
Prerequisite
馃悶 Describe the bug
As said in the title, training stops at exception.
In
mmengine/fileio/backends/local_backend.py
, line 416, I've changed into following to temp fix it.Environment
Additional information
I think it is due to removing best ckpt via multiprocessing so that it throws exception. Needs reviewing and fixing from dev side.
The text was updated successfully, but these errors were encountered: