Software Engineering Courses (SECourses)•3y ago

A short story for my niece and nephews: Tools used: 1. Stable diffusion 2. GPT-4 The main goal

A short story for my niece and nephews:

Tools used:

Stable diffusion
GPT-4

The main goals were these:

give my niece a nice present
create a coherent story
create coherent characters

All in all a rather straight forward process, just putting together several techniques I've learned in the past year.

FiveSAI•12/9/23, 4:35 AM

Even Faster SDXL Dreambooth Training Full BF16 Enabled

FiveSAI•12/9/23, 4:36 AM

Still further testing but results so far amazing

Pelolisu•12/9/23, 5:03 AM

I'm crying with 8gb vram haha

Someone.in.144p•12/9/23, 5:24 AM

same heh

FFurkan Gözükara SECourses you need to give accurate folder path and it will work

Someone.in.144p•12/9/23, 5:25 AM

it worked :DDDD, thank you so much for the help man, u savd my school project

Someone.in.144p•12/9/23, 5:25 AM

PPelolisu Click to see attachment

thewingag•12/9/23, 8:17 AM

hmm, about 100Gb lol :v

virginia•12/9/23, 11:04 AM

@Pelolisu I usually get exact dozens of images from hundreds of or even thousands of created images on the basis of trained lora, Furkan said dreambooth checkpoint is better than lora. I have not tried yet. If a object is too detailed and styled, lora results are not good, time wasting, many images are not to the standards you require.

PPflanzen die Tanzen A short story for my niece and nephews: Tools used: 1. Stable diffusion 2. ...

Furkan Gözükara SECourses•12/9/23, 11:36 AM

nice

Furkan Gözükara SECourses•12/9/23, 11:37 AM

1 picture it is merged with horse though :d

FFiveSAI Even Faster SDXL Dreambooth Training Full BF16 Enabled

Furkan Gözükara SECourses•12/9/23, 11:37 AM

huge. did you do anything special?

FFurkan Gözükara SECourses 1 picture it is merged with horse though :d

Pflanzen die TanzenOP•12/9/23, 11:55 AM

this one? The legs are under the wings

it's supposed to be like that

Pflanzen die TanzenOP•12/9/23, 11:56 AM

Or this one, which is also intended.
The protagonist turns into a unicorn and then back into a human.

FFurkan Gözükara SECourses yes i have installer

Sandstormleader•12/9/23, 12:53 PM

i think that's the one i still have installed right now, was hoping there was a working sdlx automatic1111 dreambooth at this point

Superdrew907•12/9/23, 1:09 PM

Hey jet Thank u guys so much

SSandstormleader i think that's the one i still have installed right now, was hoping there was a ...

Furkan Gözükara SECourses•12/9/23, 1:26 PM

the guy left developing

FFurkan Gözükara SECourses the guy left developing

Sandstormleader•12/9/23, 1:26 PM

sad news

Furkan Gözükara SECourses•12/9/23, 1:27 PM

ye he is working for another ai company

Furkan Gözükara SECourses•12/9/23, 1:27 PM

doing similar stuff i think

Vvirginia @Pelolisu I usually get exact dozens of images from hundreds of or even thousand...

Pelolisu•12/9/23, 3:17 PM

I think the time of 2:55 hours for my training is acceptable with an 8GB RAM GPU. However, the result of the training still didn't satisfy me, but as I'm still learning stable diffusion it may be that I haven't yet set the prompt parameters correctly. I will also try to run it on Kaggle as Furkan suggested to compare the results. I think I can use a Lion optmizer there.

Pelolisu•12/9/23, 3:19 PM

I will also do some experiments with LoRA

xlads•12/9/23, 5:09 PM

@Dr. Furkan Gözükara or anyone eles in here is there a course on how to do the deepfakes video ?

Lorelai AI•12/9/23, 7:04 PM

I'm trying to get a baseline for SDXL Lora time frames. I see a bunch of different times about train a lora in 30 minutes etc but my use case is different. Most of my SDXL Lora trainings are 12-18 hours. Now I know this has to do in part im using 768X1344. Adjusting Network Dimm seems to reduce and learning rate seems to have an impact on the time but my concern quality will be impacted if I adjust these. My Anime models are coming out great despite the time. My person models are Meh despite following the guides and using the regularization images but I believe I missed something. I'm on an RTX 4090. Any insight?

FFurkan Gözükara SECourses huge. did you do anything special?

FiveSAI•12/9/23, 7:29 PM

Yep im using pagedadamw8bit, similar results to adafactor, still testing, got about 50 more trainings to test, will try prodigy next

LLorelai AI I'm trying to get a baseline for SDXL Lora time frames. I see a bunch of differe...

FiveSAI•12/9/23, 7:30 PM

can you share your json file, also are you using the latest nvidia drivers with Sysmem fallback policy

FFiveSAI can you share your json file, also are you using the latest nvidia drivers with ...

Lorelai AI•12/9/23, 7:33 PM

That’s a good idea to check the drivers. I’ll grab the json when I have a moment

FiveSAI•12/9/23, 7:33 PM

i also have a 4090 my trainings are usually 35 mins to 2 hours depending

Lorelai AI•12/9/23, 7:34 PM

What settings are you using? I know that’s a large request. I’m all over the place even with the trainings and I’m still new.

FiveSAI•12/9/23, 7:35 PM

depends i use different settings, but usually adafactor and the extra arguments with the gradient checkpointing and sometimes xformers not always.

FFiveSAI depends i use different settings, but usually adafactor and the extra arguments ...

Lorelai AI•12/9/23, 7:36 PM

Do you think adafactor is better than AdamW (not the 8 bit)

FiveSAI•12/9/23, 7:38 PM

For me for faster training with lora i always use adamw8bit, dreambooth still testing but best is adafactor

FiveSAI•12/9/23, 7:39 PM

and for sdxl i use the noise ofset 0.0357

FiveSAI•12/9/23, 7:40 PM

also using a older version of kohya uses less vram i think 21.8.4

FiveSAI•12/9/23, 7:42 PM

also if you want me to train your data set on my system i can. so we canm compare training times

Lorelai AI•12/9/23, 7:44 PM

I’m just playing around at the moment I’m going to look at the items you mentioned. What about the basic settings like images, epochs, repeats? What I’m struggling with real people. I followed the doctors videos and did the first 2 and got just ok results.

FiveSAI•12/9/23, 7:48 PM

So far people, I use 15-30 Images, no harsh angles , usually front, 3/4 and side shots.

20 Repeats seems to be the golden amount.

Also i found using too much class images affects my training,

and usually 8 epochs, I generate samples as well, so I usually stop the training early.

FiveSAI•12/9/23, 7:48 PM

Also you may want to try using min snr gamma at the value of 10

Lorelai AI•12/9/23, 7:50 PM

How many class images roughly?

FiveSAI•12/9/23, 7:51 PM

I have my own set of 2000, but it only uses around 500

the more repeats you have the more class images it uses

Lorelai AI•12/9/23, 7:51 PM

And what SDXL base model do you use?

FiveSAI•12/9/23, 7:52 PM

Its a bit tricky so far, i use the base sdxl model 1.0, But usually loras dont work too well with other checkpoints.

So im using a custom merge of base xl juggernaught xl dreamshaper and realvis,

FiveSAI•12/9/23, 7:53 PM

but if you're only use base xl, then yeah base xl is the best

FiveSAI•12/9/23, 7:54 PM

will probably make my own tutorial soon, still testing for best settings

Lorelai AI•12/9/23, 7:54 PM

I’m setting this up that the moment thanks so much for the help. Learning rates?

FiveSAI•12/9/23, 7:54 PM

Are you training lora?

Lorelai AI•12/9/23, 7:54 PM

Yes

FiveSAI•12/9/23, 7:55 PM

If you're using adafactor

0.0001
0.0001
0.0001

for all 3

FiveSAI•12/9/23, 7:56 PM

Extra arguments

scale_parameter=False relative_step=False warmup_init=False

FiveSAI•12/9/23, 7:56 PM

you can also try one with weight decay

scale_parameter=False relative_step=False warmup_init=False weight_decay=0.01

A short story for my niece and nephews: Tools used: 1. Stable diffusion 2. GPT-4 The main goal

Similar Threads

Similar Threads

Similar Threads