Cgreat!
Fgreat
NWhy it seems to be processing double steps?
Fyou mean 1 epoch is double of training images count
Nyeah
Fthey changed
Fbucketing
Nok
Fnow classification images
Fare also included
Fso 1 training 1 class
Nit does increase the time of training?
Fwell i think now it is faster
Fbecause in lesser epoch total steps count
Fit learns
Nfine! Thats great. Thank you very much!!
CThis is alia
CThe images I use for instance
C
CI am going to put this image into the instance folder
CThis image has the following tokens: Ohwx woman black cyberpunk cenobite intricate long red hair beautiful cybernetic, realism landscape, highly detailed realistic digital painting, Art station by Alastair magnaldo adam martinakis alejandro burdisio Alison Bechdel Andreas Rocha cartoonist cyril rolando normal rockwell,beautiful face looking_at_viewer, Engine 3D vray unreal 8k uhd, Lighting studio sharp HDR ambient
C[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
CThis should theoretically not over train
Cas if I put the image in and did Ohwx woman
C2308
CI refuse to open my math books up on markov chains
Cnope nope nope
Flets see results of your experiments
Fmore prompts and more images
Fshould prevent overtrain
Cyeah
CI am generating images now
NEverything is working fine, training without issues. Is there any risk of anything updating automatically? how could i prevent that?
CI understand that there is possible a combination of token order which is not optimized for weight. But it should still have an effect
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
Fhow did you install? if you run from webui-user.bat it doesnt auto update
Forder could be affecting as well
Fnot sure though
Cit does, but you have to remember that you are most likely using too few assumptions anyway and thus overtraining is virtually impossible
CUhm I mean, thus NOT overtraining is virtually impossible, we are human
CBut no combination of the numbers
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
Will have a HIGHER weight than [5812]
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
Will have a HIGHER weight than [5812]
Fbut we also cant know what kind of affect the combination makes. also remember, after training, you have to use those prompts trained on usually to get your subject likeliness
CI want to associate looking at viewer
Cbut probably smarter to train and finetune with looking at viewer facing left
CWell of course because the images will lose weight >(
CIf I train with token
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
[5812, 49345, 2415, 2042, 10075, 30354, 269, 268, 672, 578, 28746, 890, 2266, 4190, 4950, 10075, 9833, 11, 30787, 220, 10747, 11, 4047, 6496, 12653, 4875, 12036, 11, 3683, 4429, 416, 978, 459, 958, 7842, 41476, 23197, 11277, 259, 27321, 31341, 47983, 8746, 271, 952, 33629, 1355, 354, 12381, 33728, 371, 5374, 64, 16251, 396, 269, 2417, 346, 686, 11993, 3487, 3881, 4053, 11, 40544, 4135, 1986, 2045, 62, 265, 62, 1177, 263, 11, 7117, 513, 35, 410, 2433, 22865, 807, 74, 21480, 67, 11, 43150, 8034, 7786, 33675, 25237, 220, 220, 198]
Cand then use token 5812, it will not be as affected
Cmakes perfect sense
NOk! i think its not risk then
Fyep
Funless there is some auto update command inside it. what you have inside that file