Params

MODEL_NAME="stable-diffusion-2-1-base"; 
LR = "constant"
LEARNING_RATE="1.e-4"
LEARNING_RATE_TEXT="5e-5"
TRAINING_STEPS="1000"

Thoughts

UNet: 1.e-4, Test: 5e-5, Training steps: 1000 - 4 Images
- Unet 0.75-1 is sort of okay, but still its hit and miss, not very impressed with these params.
- Beyond 1.3 it looked so horrible, I stopped the whole test :D

Original Images:

Unet: 0, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0.25, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 0.25

a closeup portrait of sjqw1 woman

Unet: 0.25, Text: 0.25

a closeup portrait of sjqw1 woman

Unet: 0.5, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 0.5

a closeup portrait of sjqw1 woman

Unet: 0.5, Text: 0.5

a closeup portrait of sjqw1 woman

Unet: 0.5, Text: 0.25

a closeup portrait of sjqw1 woman

Unet: 0.25, Text: 0.5

a closeup portrait of sjqw1 woman

Unet: 0.75, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 0.75

a closeup portrait of sjqw1 woman

Unet: 0.75, Text: 0.75

a closeup portrait of sjqw1 woman

Unet: 0.75, Text: 0.5

a closeup portrait of sjqw1 woman

Unet: 0.5, Text: 0.75

a closeup portrait of sjqw1 woman

Unet: 1, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 1

a closeup portrait of sjqw1 woman

Unet: 1, Text: 1

a closeup portrait of sjqw1 woman

Unet: 1, Text: 0.75

a closeup portrait of sjqw1 woman

Unet: 0.75, Text: 1

a closeup portrait of sjqw1 woman

Unet: 1.3, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 1.3

a closeup portrait of sjqw1 woman

Unet: 1.3, Text: 1.3

a closeup portrait of sjqw1 woman

Unet: 1.3, Text: 1

a closeup portrait of sjqw1 woman

Unet: 1, Text: 1.3

a closeup portrait of sjqw1 woman

Unet: 1.5, Text: 0

a closeup portrait of sjqw1 woman

Unet: 0, Text: 1.5

a closeup portrait of sjqw1 woman

Unet: 1.5, Text: 1.5

a closeup portrait of sjqw1 woman

Unet: 1.5, Text: 1.3

a closeup portrait of sjqw1 woman

Unet: 1.3, Text: 1.5

a closeup portrait of sjqw1 woman