Software Engineering Courses (SECourses)•3y ago

I am training right now: Theres an issue with sanity prompt. I write "a photo of ---- by Tomer Hanuk

I am training right now: Theres an issue with sanity prompt. I write "a photo of ---- by Tomer Hanuka. But, as soon as 640 steps, the style dissapears. And it gives me realistic photos instead. The model dont look overtraining, face are still not right. I have 32 sample images: portraits and some mid a full body..

Furkan Gözükara SECourses•1/24/23, 12:19 PM

Nvidia may also make 24 hb rtx 4060 that would be also a good upgrade

NNandux I am training right now: Theres an issue with sanity prompt. I write "a photo of...

Furkan Gözükara SECourses•1/24/23, 12:20 PM

Can you try half learning rate

FFurkan Gözükara SECourses Nvidia may also make 24 hb rtx 4060 that would be also a good upgrade

NanduxOP•1/24/23, 12:20 PM

It would be nice and more affordable for sure

Furkan Gözükara SECourses•1/24/23, 12:20 PM

Half learning rate can help a lot

Furkan Gözükara SECourses•1/24/23, 12:20 PM

0.000001

NanduxOP•1/24/23, 12:20 PM

Num batches each epoch = 32
Num Epochs = 100
Batch Size Per Device = 2
Gradient Accumulation steps = 2
Total train batch size (w. parallel, distributed & accumulation) = 4
Text Encoder Epochs: 75
Total optimization steps = 3200
Total training steps = 6400
Resuming from checkpoint: False
First resume epoch: 0
First resume step: 0
Lora: False, Adam: True, Prec: bf16
Gradient Checkpointing: False
EMA: True
UNET: True
Freeze CLIP Normalization Layers: False
LR: 8e-06

Furkan Gözükara SECourses•1/24/23, 12:20 PM

You are using 0.000008

NanduxOP•1/24/23, 12:20 PM

ok! i am going to try it

Furkan Gözükara SECourses•1/24/23, 12:20 PM

Try half

Furkan Gözükara SECourses•1/24/23, 12:20 PM

If still try half again