Software Engineering Courses (SECourses)•3y ago

any special argument ? keywords ? etc to be used ? Or ohwx / bbuk woman and man keys will be taken f

any special argument ? keywords ? etc to be used ? Or ohwx / bbuk woman and man keys will be taken from lora automatically ? I dont want to create a ckpt/safetensor , it will limit my usuage to be used with other good models. I will see if thre is possibility of merging them into one lora for two faces. Or @Dr. Furkan Gözükara is it possible to traing lora for two faces with captions ? That will be easy i guess ?

FFiveSAI Yep dreambooth, but suprisingly the settings somewhat work with lora too

Benjamin•12/6/23, 8:05 AM

sounds interesting, do you mind sharing your training settings?

FFurkan Gözükara SECourses comfy ui manager

740•12/6/23, 8:42 AM

They are not showing Up in comfyui manager

7740 missing nodes doesnt appers in manager

740•12/6/23, 8:42 AM

I have already installed all missing custom nodes from that tab, but still missing these

hozaggers•12/6/23, 10:24 AM

I am trying to follow the tutorial on https://www.patreon.com/posts/kohya-sdxl-lora-88397937
when I try to 'start training' I get an output
09:59:46-038102 INFO Start training Dreambooth...
09:59:46-039759 INFO Image folder does not exist

not clear to me which Image folder it doesn't like... here is my inputs and my kaggle file... what am I doing wrong?

Deleted User•12/6/23, 10:28 AM

Click on prepare training data firts then copy info to folder

hozaggers•12/6/23, 10:28 AM

aha did the trick thank you

Deleted User•12/6/23, 10:41 AM

You are welcome

hozaggers•12/6/23, 11:06 AM

does anyone know what might be causing the training to fail?

message.txt2.53KB

Bbaloooo My workflow: 1. I play in GUI 2. Spy with the GIT package above 3. Do iterative ...

andrsoe•12/6/23, 11:09 AM

wait so, you recreate your UI results with the json api call with 40% succesrate ? SUS

Hhozaggers does anyone know what might be causing the training to fail?

Deleted User•12/6/23, 11:28 AM

The number of images must take at least 30 to catche

Deleted User•12/6/23, 11:28 AM

Your catche must be finished inside 30 minutes

Deleted User•12/6/23, 11:29 AM

Your training killed in 1577 images catched so resuce your training dataset images to 1550

Hhozaggers does anyone know what might be causing the training to fail?

Deleted User•12/6/23, 11:30 AM

How many training inages in your data set

Elisoli•12/6/23, 11:30 AM

on magic animate i get this error and it stuck like this :

pipeline_animation.py:624: FutureWarning: Accessing config attribute in_channelsin_channels directly via 'UNet3DConditionModel' object attribute is deprecated. Please access 'in_channels' over 'UNet3DConditionModel's config object instead, e.g. 'unet.config.in_channels'.
num_channels_latents = self.unet.in_channels
0%| | 0/25 [00:00<?, ?it/s]

Deleted User•12/6/23, 11:33 AM

During extract lora from my checkpoint in kaggle i got error out of memory how to solve the problem and use gpu during the process

DDeleted User How many training inages in your data set

hozaggers•12/6/23, 11:39 AM

I have 174 images in training dataset, although the regularization I think has 5200? if I re-run will it continue to cache them until it caches them all? or does it start caching from the start each time?

Hhozaggers I have 174 images in training dataset, although the regularization I think has 5...

Deleted User•12/6/23, 11:40 AM

It will start catche from 0

DDeleted User It will start catche from 0

hozaggers•12/6/23, 11:42 AM

the error says 1577/5374 which I think makes sense. 5200 regularization images + 174 of my own training images. not anything to do with my internet upload speed since everything is already uploaded to Kaggle? it still seems quite a low number to timeout at (not even going through all the regularization images)

Deleted User•12/6/23, 11:42 AM

Reduce your training images to 50

Hhozaggers the error says 1577/5374 which I think makes sense. 5200 regularization images +...

Deleted User•12/6/23, 11:43 AM

It has nothing to do with regulationsimages

Deleted User•12/6/23, 11:43 AM

Just reduce your training images to 50

Deleted User•12/6/23, 11:44 AM

If you reduce to 50 images it will do the job for you

Bbaloooo My workflow: 1. I play in GUI 2. Spy with the GIT package above 3. Do iterative ...

Benjamin•12/6/23, 11:49 AM

what analysis does your script?

Deleted User•12/6/23, 11:57 AM

Anyone will help me with

Deleted User•12/6/23, 11:58 AM

When i do extract lora from my dreambooth checkpoint error comes that your notebook allocates more memory than available it is not utilising gpu

Aandrsoe wait so, you recreate your UI results with the json api call with 40% succesrate...

baloooo•12/6/23, 12:09 PM

1st. I create the payload with UI
2nd. I work on improving the payload, iteratively as described, by adjusting parameters: prompt, noise, CFG, additional scripts, etc.
The project workscope. >100 tuned models; >20 JSON payloads.
So that one payload might work well for a particular model, but poorly for another model.
Success rate (average 40%) - for all models, for all payloads. We generate thousands of images and remove 60%, manually. SD (stand dev) is quite high though. Can be improved, but I don't care that much.

BBenjamin what analysis does your script?

baloooo•12/6/23, 12:10 PM

pretty basic statistical analysis: tables with mean and sd

DDeleted User If you reduce to 50 images it will do the job for you

hozaggers•12/6/23, 12:30 PM

hmm reduced to 50 training images, it seems to be timing out at a similar point

message.txt2.42KB

DR.SivaOP•12/6/23, 12:46 PM

i used only 13

DR.SivaOP•12/6/23, 12:47 PM

even at 18 it was timing out.. i was training lora sdxl

DR.SivaOP•12/6/23, 12:47 PM

also check in the argument in kohyass to remove "text_encoder_learner" something

hozaggers•12/6/23, 1:15 PM

should I try reduce my training images further? (from 50 to 10?) and uncheck "text_encoder_learner"? (where do I find this?) feel like I am missing something fundamental...

Hhozaggers should I try reduce my training images further? (from 50 to 10?) and uncheck "t...

Deleted User•12/6/23, 1:30 PM

13 with 1024,1024 working good

Hhozaggers should I try reduce my training images further? (from 50 to 10?) and uncheck "t...

Deleted User•12/6/23, 1:31 PM

As you are using 512,512 maybe more images around 15 to 20 will also work

Bbaloooo 1st. I create the payload with UI 2nd. I work on improving the payload, iterativ...

andrsoe•12/6/23, 1:31 PM

Im really struggling to explain my self, for that i am sorry, but:
-You describe the workflow, super nice!
-But are you saying if you could just use the UI you would hit 100% correct generations, and you are "fighting the results you are getting from UI vs the json payload"?

DDeleted User As you are using 512,512 maybe more images around 15 to 20 will also work

hozaggers•12/6/23, 1:32 PM

thanks, is there anyway to increase the timeout or is this a limitation by Kaggle? ideally I want to use all my training data

Hhozaggers thanks, is there anyway to increase the timeout or is this a limitation by Kaggl...

Deleted User•12/6/23, 1:32 PM

It’s bug maybe they will solve the problem in near future

Hhozaggers thanks, is there anyway to increase the timeout or is this a limitation by Kaggl...

Deleted User•12/6/23, 1:33 PM

You are doing 512,512 training

hozaggers•12/6/23, 1:34 PM

yes I am doing 512x512

Hhozaggers yes I am doing 512x512