require a missing '.experiment' even for the first time running

[FrankYJY]

Describe the issue: When I tried HPO Quickstart with PyTorch, I stuck in Error: ENOENT: no such file or directory, open 'C:\Users\14048\nni-experiments.experiment' in this snapshot, you can see that it keeps asking for "user/nni-experiments/.experiment" repeatedly (I understand the repeating should be the refreshing request to the backend data) after I tried to put a blank .experiment manually, I found out that it(from browser log as the screenshot above) requires a json file, and some information is required, without which I cannot run. It requires that file even for the first time running.

what nni-experiment looks like, I tried to delete the whole folder but not work

Environment:

• NNI version: 3.0
• Training service (local|remote|pai|aml|etc): local
• Client OS: windows anaconda
• Server OS (for remote mode only):
• Python version: 3.10.13
• PyTorch/TensorFlow version: 2.1.0
• Is conda/virtualenv/venv used?: conda
• Is running in Docker?: Windows Anaconda

Configuration: everything default

Log message:

• nnimanager.log:
• dispatcher.log:
• nnictl stdout and stderr:

How to reproduce it?:

[ducbueno]

Anyone found a solution to this? I'm having the same issue

[ducbueno]

I think I found the solution to this. It was given as an answer to issue https://github.com/microsoft/nni/issues/5523. Apparently it is a problem with newer versions of the filelock package. To me the problem was gone when I downgraded filelock with pip install filelock==3.10.