Skip to content

Latest commit

 

History

History
21 lines (12 loc) · 1010 Bytes

README.md

File metadata and controls

21 lines (12 loc) · 1010 Bytes

To generate images using Stable Diffusion v1-4, you can run the following:

OBJ_VALUES=("a mountain landscape" "a flamingo" "a dragonfly" "dandelion" "a sunflower" "a rocket" "moon" "a snail" "an eagle" "zebra" "chess pawn" "a pineapple" "a spider web" "a waffle cone" "a cat" "a chair" "a donut" "otter" "pebbles on a beach" "teddy bear")
BG_VALUES=("silhouette of a dog" "a candy cane"  "a helicopter" "fireworks" "a lemon" "a cactus" "cookie" "a cinnamon roll" "an airplane" "barcode" "bottle cap" "a beehive" "a bicycle wheel" "a volcano" "a dog" "an avocado" "a map" "duck" "a turtle" "panda")

METHODS=("and" "avg" "sd_ab" "sd_ba" "or" "sd_ab_or" "sd_ba_or" "sd_a" "sd_b")

TASK_ID=0
OBJ=${OBJ_VALUES[$TASK_ID]}
BG=${BG_VALUES[$TASK_ID]}

METHOD=${METHODS[$1]}

echo "Running job with METHOD=${METHOD}, OBJ=${OBJ}, BG=${BG}"

python clip_eval.py --num_inference_steps 1000 --batch_size 1 --method "${METHOD}" --obj "${OBJ}" --bg "${BG}"