Software Engineering Courses (SECourses)•4mo ago

why are you training checkpoint on a character, that's what lora is for 😉

why are you training checkpoint on a character, that's what lora is for

Jimi•9/2/25, 9:23 AM

i used the gonzaLomo XL model

JJonkoXL why are you training checkpoint on a character, that's what lora is for 😉

Jimi•9/2/25, 9:24 AM

i tried lora in flux but dream boot is way better result compare that

Jimi•9/2/25, 9:24 AM

have't tried SDXL lora yet

JonkoXLOP•9/2/25, 9:27 AM

Somehow I got good results with SDXL Lora trained on this (Pony) checkpoint. (It's NSFW, but seems to be giving good results)

JonkoXLOP•9/2/25, 9:27 AM

https://civitai.com/models/643732/animatedponyreal

Jimi•9/2/25, 9:31 AM

do you use Kohya for SDXL lora?

JonkoXLOP•9/2/25, 9:31 AM

koyha_ss gui

JonkoXLOP•9/2/25, 9:31 AM

yes

Jimi•9/2/25, 9:31 AM

do you have any setting to share

JonkoXLOP•9/2/25, 9:31 AM

those are on my old pc

JonkoXLOP•9/2/25, 9:31 AM

Default works I believe

Jimi•9/2/25, 9:31 AM

ohh

JonkoXLOP•9/2/25, 9:31 AM

try to aim for 1600-3000 steps

JonkoXLOP•9/2/25, 9:31 AM

at least

JonkoXLOP•9/2/25, 9:32 AM

and enable sampling

JonkoXLOP•9/2/25, 9:32 AM

at 1000+ it should resemble whatever you wanted to train

JonkoXLOP•9/2/25, 9:32 AM

and at 1800 it should be decent enough

Jimi•9/2/25, 9:32 AM

got it

Jimi•9/2/25, 9:32 AM

will try that

JonkoXLOP•9/2/25, 9:32 AM

what captions do you have?

JonkoXLOP•9/2/25, 9:32 AM

because less is often better

Jimi•9/2/25, 9:33 AM

i have't use any captions

Jimi•9/2/25, 9:33 AM

should i have to use?

JonkoXLOP•9/2/25, 9:33 AM

try using captions

Just describe what you dont want to have in the image

Jimi•9/2/25, 9:33 AM

okay

JonkoXLOP•9/2/25, 9:34 AM

so for a woman in a park with a bench and a tree, and some dog
-> myL0r4K3yw0rd, park, bench, tree, dog

JonkoXLOP•9/2/25, 9:34 AM

dont mention the woman or her body features , you can mention her clothes however

Jimi•9/2/25, 9:34 AM

can i use khoya captions?

Jimi•9/2/25, 9:34 AM

i mean blip one

JonkoXLOP•9/2/25, 9:35 AM

yes, but you should manually remove the ones refering to the likeness of the woman

Jimi•9/2/25, 9:35 AM

okayy

JonkoXLOP•9/2/25, 9:35 AM

so " woman, young face, freckles, curvy" you should never want in the captions for example (for a character)

Jimi•9/2/25, 9:35 AM

it will be generic

JonkoXLOP•9/2/25, 9:36 AM

Just describe as captions what you dont want to have in the image

Jimi•9/2/25, 9:36 AM

ya got it

Jimi•9/2/25, 9:36 AM

Thanks

JonkoXLOP•9/2/25, 9:37 AM

please share your results afterwards

JonkoXLOP•9/2/25, 9:37 AM

and definitely use sampling every 100 steps

Jimi•9/2/25, 9:39 AM

okay will do that

Jimi•9/2/25, 9:39 AM

i am trying diffrent dataset in diffrent model. i will try wan next then qwen

JonkoXLOP•9/2/25, 9:40 AM

wan 2.1 or 2.2?

JonkoXLOP•9/2/25, 9:40 AM

2.1 may be easier

Jimi•9/2/25, 9:42 AM

2.2

Jimi•9/2/25, 9:42 AM

but i think i will take long time

Jimi•9/2/25, 9:42 AM

i will use diffusion pipe

JonkoXLOP•9/2/25, 9:45 AM

dont go with 2.2

JonkoXLOP•9/2/25, 9:45 AM

do 2.1 first

JonkoXLOP•9/2/25, 9:45 AM

it takes less time (1 model) and grasps things quicker

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•9/2/25, 9:46 AM

@Dr. Furkan Gözükara when using swarmui inside massed, is it enough to just run update, then swarm? did u already set up the comfyui backend there right?

why are you training checkpoint on a character, that's what lora is for 😉

Similar Threads

Similar Threads

Similar Threads