|
num_frames = 1 |
|
fps = 1 |
|
image_size = (256, 256) |
|
|
|
|
|
model = dict( |
|
type="DiT-XL/2", |
|
no_temporal_pos_emb=True, |
|
condition="text", |
|
from_pretrained="PRETRAINED_MODEL", |
|
) |
|
vae = dict( |
|
type="VideoAutoencoderKL", |
|
from_pretrained="stabilityai/sd-vae-ft-ema", |
|
) |
|
text_encoder = dict( |
|
type="clip", |
|
from_pretrained="openai/clip-vit-base-patch32", |
|
model_max_length=77, |
|
) |
|
scheduler = dict( |
|
type="dpm-solver", |
|
num_sampling_steps=20, |
|
cfg_scale=4.0, |
|
) |
|
dtype = "fp16" |
|
|
|
|
|
batch_size = 2 |
|
seed = 42 |
|
prompt_path = "./assets/texts/imagenet_labels.txt" |
|
save_dir = "./outputs/samples/" |
|
|
