configs/with_mask_sample.yaml

# path config:
ckpt: "pretrained/ShowMaker.pt"
pretrained_model_path: "pretrained/OpenCLIP-ViT-H-14"
input_path: 'input/i2v/Planet_hits_earth.png'
save_path: "results/mask_no_ref/"

# model config: 
model: UNet
num_frames: 16
# image_size: [320, 512]
image_size: [240, 560]

# model speedup
use_fp16: True
enable_xformers_memory_efficient_attention: True

# sample config:
seed: 3407
cfg_scale: 8.0
sample_method: 'ddim'
num_sampling_steps: 100
text_prompt: [
              # "Cinematic photograph. View of piloting aaero.",
              # "A fish swims past an oriental woman.",
              # "A big drop of water falls on a rose petal.",
              # "Underwater environment cosmetic bottles.".
              "Planet hits earth.",
              ]
additional_prompt: ""
negative_prompt: ""
do_classifier_free_guidance: True
mask_type: "first1"
use_mask: True