Multiple safetensors output

[CrisOrlandoBR]

Hello, I am new here and I started to use Kohya yesterday
My data set was big (almost 700 images) and I was able to make it work (almost 15h using a dual xeon x99 2011-v3 18/36 and a 4090 GPU (gonna try using my Ryzen 9 with a 3080TI too)

But here is the thing = It created a lot of safetensor files
There is one with the output name and it also created more 49 files with the same size (NAMEMODEL-000001 to NAMEMODEL-000049
All files have the same size (220mb) and I used a big pretrained model (7gb)
Is this ok ? Do I have to merge those files or they are the same ? If so, is there a script to merge a large number of safetensors ? I checked some extensions and some github projects with safetensor mergers, and they are usually made to merge just 1 to 3 files...

And in image folder I have a lot of NPZ files

Do I have to delete them ?

Tks

[JimmyAlan]

I am encountering the same. I am wondering if you found anything out?

[CommieDog]

Those sound like the model results at different epochs of training. Each one of those files is the full model, just at different stages of training. Which file is the right one depends on the training parameters and personal preference; earlier epochs may be undertrained (not close enough to what you're trying to train) but later epochs may be overtrained (too close to the data used in training and inflexible). Picking the epoch with the right amount of training is an art, not a science.