Unuseable .bk2 files across the dataset

[hyruuk]

I've found unuseable .bk2 files across the dataset. The two reasons I identified were :

• fakereps : Repetitions stopped early (max score below 200)
• emptyfiles : Corrupted files (lead to "movie couldn't be loaded" errors when replayed)

I attached two text files containing the culprit's list :
home_emptyfiles.txt
home_fakereps.txt

[yibeichan]

hello, I just used retro to process those bk2 files and found most of sub-01's training bk2 files have error: Error in main execution: No romfiles found for game: CLS these failed files overlapped with your [home_fakereps.txt](https://github.com/courtois-neuromod/shinobi_training/files/8007392/home_fakereps.txt) list
and some sub-06 files have could not load movie error:

'shinobi_training/sub-06/ses-016/beh/sub-06_ses-016_task-shinobi_level-1_rep-001_gamedata.bk2'
'shinobi_training/sub-06/ses-014/beh/sub-06_ses-014_task-shinobi_level-4_rep-005_gamedata.bk2'
'shinobi_training/sub-06/ses-010/beh/sub-06_ses-010_task-shinobi_level-4_rep-010_gamedata.bk2'
'shinobi_training/sub-06/ses-013/beh/sub-06_ses-013_task-shinobi_level-4_rep-001_gamedata.bk2'
'shinobi_training/sub-06/ses-015/beh/sub-06_ses-015_task-shinobi_level-1_rep-001_gamedata.bk2'
'shinobi_training/sub-06/ses-002/beh/sub-06_ses-002_task-shinobi_level-1_rep-004_gamedata.bk2'

@hyruuk just want to double check with you, so those files/sessions were removed from your final analyses, right? no way to recover them?

[hyruuk]

Hi !

I remember encountering this error too, I think this is caused by the way .bk2 were saved during the training phase of shinobi which might have omitted to save the proper game name. The workaround I found was to pass the game name explicitely to the function I used to read the replay. You can have a look there, its in a separate branch hence why it's not easy to find yet : https://github.com/courtois-neuromod/bids_loader/blob/dev_replays/bids_loader/stimuli/game.py

There are also a few functions that could help your exploration, have a look :)

About the fakereps and emptyfiles yes, they were removed from the analysis. There is nothing to be recovered, they were just created abnormally for two different reasons. I think they should be deleted from the dataset, thanks for getting the issue to my attention.

glhf

[yibeichan]

hey, thank you for these functions! I didn't know the sound/audio was an option to be extracted as well. the plotting function is nice too!