# machine-learning
- h
hecko
11/17/2021, 10:32 PMbecause the neural network never has neurons added or removed, only adjusted - u
user
11/17/2021, 10:59 PMohhhhhhh - u
user
11/17/2021, 11:00 PMwhy are there more than 1 checkpoints? is it to prevent losing it if your pc crashes - c
Couch
11/17/2021, 11:00 PMdifferent runs - c
Couch
11/17/2021, 11:00 PMtrain.py only saves once - u
user
11/17/2021, 11:04 PMso whats the difference between all the files - u
user
11/17/2021, 11:04 PMwhich file should i use for the model - c
Couch
11/17/2021, 11:04 PMit tells you the file it saves to when you finish running train.py - u
user
11/17/2021, 11:05 PMno it didnt - u
user
11/17/2021, 11:05 PMit just did nothing - c
Couch
11/17/2021, 11:06 PMim going to shit - c
Couch
11/17/2021, 11:06 PMok sort the checkpoint folder by date - c
Couch
11/17/2021, 11:06 PMalso 1.6k epochs is a lot for the dataset youre doing - u
user
11/17/2021, 11:07 PMtheres 2 files checkpoint_0 and checkpoint_1000 - u
user
11/17/2021, 11:07 PMik it was just a test - c
Couch
11/17/2021, 11:07 PM_1000 - u
user
11/17/2021, 11:07 PMso wait - u
user
11/17/2021, 11:07 PM_1000 is finished? - c
Couch
11/17/2021, 11:07 PMyes - u
user
11/17/2021, 11:07 PMwhy did it not tell me what file it did when it finished - c
Couch
11/17/2021, 11:07 PMpython estupido - u
user
11/17/2021, 11:07 PMso its not broken? - u
user
11/17/2021, 11:09 PMwell i assume not - u
user
11/17/2021, 11:09 PMi havent tried the model yet - u
user
11/17/2021, 11:19 PMthe model sounds like it's having a stroke it failed completely, is this because i used way too many epochs? @User - c
Couch
11/17/2021, 11:19 PMyou overfitted (too many epochs for 37 file dataset) - c
Couch
11/17/2021, 11:19 PMwhats your graph look like - u
user
11/17/2021, 11:20 PMabsolute shit - c
Couch
11/17/2021, 11:20 PMyou overfitted - u
user
11/17/2021, 11:20 PMhow do you calculate what is a good amount to do