I want to train my own model basic training process has shortcomings

[oztrkoguz]

Are there sep_trainlist.txt, sep_testlist.txt sample files in model/dataset.py, how should I place the visuals for training?
@98mxr

[98mxr]

Simply put you need to place the image pairs in folders and then create the /vimeo_septuplet/sequences folder in the root directory to contain them

like

vimeo_septuplet
├── sequences
│   ├── 1
│   ├── 2
...
│   └── 10000

imgs in folder 1, 2... 10000

sep_trainlist.txt is only need
1 2 10000

It's enough that sep_testlist.txt is an empty file; in fact, the test-related code never works properly.
I know I'm not describing it very well, so if there's still some confusion, please let me know and I'll upload the relevant content later.

[oztrkoguz]

This is how I set everything up
I ran train_nb.py and waited for a certain amount of time before the training started and it ended without giving any errors.
@98mxr

[98mxr]

This is correct, just replace the content with your dataset