tried to make a pose lora that can show 2 characters instead of 1
v1 triggers
sitting on person, full body, crossed legs = works, the rest not so much
from side
profile
straight-on
ass focus, from behind
solo focus, from below
Description
trained on 21 images, girl on top only
this is the 2nd ver actually. the first had more tags like "sitting on boy" and "sitting on girl". it was too many tags so retrained without but i removed too many tags. i removed "all fours". i should have kept that.
Training Params
{
"engine": "kohya",
"unetLR": 0.0009,
"clipSkip": 1,
"loraType": "lora",
"keepTokens": 1,
"networkDim": 8,
"numRepeats": 10,
"resolution": 1024,
"lrScheduler": "cosine_with_restarts",
"minSnrGamma": 5,
"noiseOffset": 0.3,
"targetSteps": 1050,
"enableBucket": true,
"networkAlpha": 16,
"optimizerType": "Adafactor",
"textEncoderLR": 0.00001,
"maxTrainEpochs": 20,
"shuffleCaption": true,
"trainBatchSize": 4,
"flipAugmentation": true,
"lrSchedulerNumCycles": 3
}