Software Engineering Courses (SECourses)•15mo ago

its scary what the future holds lol

Furkan Gözükara SECourses•10/21/24, 10:42 PM

haha background has bunch of tourits

0213-photograph_of_ohwx_man_as_an_Ottoman_Adm-best_modelsFP8_256_img_BS_1-000070-788251335.png

DDazzastrous [4090]its scary what the future holds lol

Furkan Gözükara SECourses•10/21/24, 10:42 PM

ye we definitely wont be able to differentiate

Dazzastrous [4090]OP•10/21/24, 10:43 PM

might be merchants

Dazzastrous [4090]OP•10/21/24, 10:43 PM

lol

Furkan Gözükara SECourses•10/21/24, 10:43 PM

haha this one epic

0215-ohwx_man_poster_design_for_a_theatre_pr-best_modelsFP8_256_img_BS_1-000070-788251371.png

Dazzastrous [4090]OP•10/21/24, 10:43 PM

coming to trade with the captain

Furkan Gözükara SECourses•10/21/24, 10:44 PM

i can never return back sdxl

Dazzastrous [4090]OP•10/21/24, 10:44 PM

liking the expressions

DDazzastrous [4090]liking the expressions

Furkan Gözükara SECourses•10/21/24, 10:44 PM

ye i tried to add

Furkan Gözükara SECourses•10/21/24, 10:44 PM

Furkan Gözükara SECourses•10/21/24, 10:45 PM

by the way ai also generates accurate expression according to prompt

Furkan Gözükara SECourses•10/21/24, 10:45 PM

as long as you have them in training dataset

Ddsienra @Dr. Furkan Gözükara Update https://huggingface.co/nyanko7/flux-dev-de-distill L...

ar.1•10/21/24, 10:45 PM

how are you inferencing that ? I saw in the model card they say it is not compatible with the normal pipeline and they provide an inference script, so does it work like normal flux is or is it different ?

Furkan Gözükara SECourses•10/21/24, 10:45 PM

0218-photograph_of_smiling_ohwx_man_riding_a-best_modelsFP8_256_img_BS_1-000070-788251399.png

Dazzastrous [4090]OP•10/21/24, 10:46 PM

do u have to tell the ai this is a pic of ohxman smiling , angry , etc?

Dazzastrous [4090]OP•10/21/24, 10:46 PM

in the dataset

Furkan Gözükara SECourses•10/21/24, 10:46 PM

0221-World_of_Warcraft_game_engine_3d_cgi_ren-best_modelsFP8_256_img_BS_1-000070-788251431.png

DDazzastrous [4090]do u have to tell the ai this is a pic of ohxman smiling , angry , etc?

Furkan Gözükara SECourses•10/21/24, 10:46 PM

i only told smiling

Dazzastrous [4090]OP•10/21/24, 10:47 PM

lol pixar

Furkan Gözükara SECourses•10/21/24, 10:47 PM

this expression by ai

Furkan Gözükara SECourses•10/21/24, 10:47 PM

so prompts were test prompts in zip file

Furkan Gözükara SECourses•10/21/24, 10:47 PM

smiling added

Furkan Gözükara SECourses•10/21/24, 10:47 PM

generated both cases

Dazzastrous [4090]OP•10/21/24, 10:47 PM

very good

Furkan Gözükara SECourses•10/21/24, 10:47 PM

but didnt add any other expression like angry etc

Furkan Gözükara SECourses•10/21/24, 10:48 PM

for example this is purely ai

0222-ohwx_man_poster_design_for_a_theatre_pr-best_modelsFP8_256_img_BS_1-000070-788251438.png

Furkan Gözükara SECourses•10/21/24, 10:48 PM

but in training dataset i had tried to add such ones

Dazzastrous [4090]OP•10/21/24, 10:48 PM

very nice

Dazzastrous [4090]OP•10/21/24, 10:49 PM

michael hutchence

Furkan Gözükara SECourses•10/21/24, 10:50 PM

haha i have no such ahircut

0234-World_of_Warcraft_game_engine_3d_cgi_ren-best_modelsFP8_256_img_BS_1-000070-788251543.png

Dazzastrous [4090]OP•10/21/24, 10:50 PM

just have the armour huh doc?

Furkan Gözükara SECourses•10/21/24, 10:50 PM

almost every image is good

Furkan Gözükara SECourses•10/21/24, 10:51 PM

with SDXL i was scrolling 100s to get good ones

Dazzastrous [4090]OP•10/21/24, 10:51 PM

yeh these are superb quality

Furkan Gözükara SECourses•10/21/24, 10:51 PM

0241-World_of_Warcraft_game_engine_3d_cgi_ren-best_modelsFP8_256_img_BS_1-000070-788251610.png

Furkan Gözükara SECourses•10/21/24, 10:52 PM

people didnt bleed

0244-photograph_of_ohwx_man_wearing_an_amazin-best_modelsFP8_256_img_BS_1-000070-788251629.png

Aar.1 how are you inferencing that ? I saw in the model card they say it is not compat...

dsienra•10/21/24, 10:52 PM

Is compatible with Forge, you just set the cfg scale at 3.5 and it works, The loras I trained with this model work on de-distilled model and in the normal model, In the normal model is more inconsistent you have to play the seed lottery, sometimes you get perfect resemblance sometimes no so much, the advantage of training on this model is that you can train many people in one lora, something that is imposible on regular flux and it learns concepts much better without bleeding

Dazzastrous [4090]OP•10/21/24, 10:52 PM

heheheh

Dazzastrous [4090]OP•10/21/24, 10:52 PM

brilliant

Dazzastrous [4090]OP•10/21/24, 10:52 PM

i like the quirky ones

Dazzastrous [4090]OP•10/21/24, 10:53 PM

eccentric

Furkan Gözükara SECourses•10/21/24, 10:53 PM

wow people looks accurate