configs/sample.yaml

# ckpt
ckpt: /mnt/lustre/maxin/work/animation/animation-v6/results_qt_003-UNet-pixabaylaion-Xfor-Gc-320_512_0303000.pt
save_img_path: "./sample_videos/"

# pretrained_model_path: "/mnt/hwfile/gcc/maxin/work/pretrained/stable-diffusion-v1-4/"
pretrained_model_path: "maxin-cn/Cinemo"

# model config: 
model: UNet
video_length: 15
image_size: [320, 512]
# beta schedule
beta_start: 0.0001
beta_end: 0.02
beta_schedule: "linear"

# model speedup
use_compile: False
use_fp16: True

# sample config:
seed:
run_time: 0
use_dct: False
guidance_scale: 7.5 #
motion_bucket_id: 95 # [0-19] The larger the value, the stronger the motion intensity
sample_method: 'DDIM'
num_sampling_steps: 50
enable_vae_temporal_decoder: True
image_prompts: [
              ['aircraft.jpg', 'aircrafts flying'],
              ['car.jpg' ,"car moving"],
              ['fireworks.jpg', 'fireworks'],
              ['flowers.jpg', 'flowers swaying'],
              ['forest.jpg', 'people walking'],
              ['shark_unwater.jpg', 'shark falling into the sea'],
              ]