-
Notifications
You must be signed in to change notification settings - Fork 3
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update metrics to the ones used in the thesis #12
Comments
Looking at the thesis logs, I finished the evaluation at around this time
|
The checkpoint files for ipython notebooks might be a good option. Although they are apparently only updated on manual save. There are several files with timestamps of "Sep XX" there, so maybe I can recover things after all? |
I also checked my backups, but because I had issues with Time Machine, I had no backups between March 2019 and May 2020. So it is checkpoints or bust. |
This is what I found from the checkpoints. I think most of them are fairly accurate, and if not, we can fix them now. This is a partial fix for MobilityNet/mobilitynet.github.io#12
Checked in the checkpointed files; now we just have to go through them one at a time and see which version to keep |
There were also some files which were not checked in, and so were not modified in the reset. Given that they are not modified, and I ran the same metrics calculation for the baseline and the analysed master, these can provide an additional clue to the final code that I used.
|
Which were not checked in before, and probably represent the most recent version of the metrics. This is another partial fix for MobilityNet/mobilitynet.github.io#12
But was sticking around in the directory, and we don't want to lose either. This may help with MobilityNet/mobilitynet.github.io#12 but I am not going to commit it right away
For the analysis_master files (which were not reset), the timestamps for the raw notebook and the checkpoint match consistently. This is a pretty positive indication that we were in fact checkpointing regularly. And since the checkpoints are from late Sept and we checked in the figures on Oct 2, at worst, we have lost the changes between Sep 25 and Oct 2. So the planned resolution (@singhish) is:
If they don't match, file an issue with the details for me to investigate.
|
As a concrete example of
So we pick the checkpoint implementation all through |
While fixing #11
I inadvertently reset from head, which reverted all my local changes.
I believe I had implemented better versions of multiple metrics in different notebooks.
Need to see if I can recover them.
The text was updated successfully, but these errors were encountered: