from diffusers import Transformer2DModel
model = Transformer2DModel(
    activation_fn="gelu-approximate",
    attention_bias=True,
    attention_head_dim=64,
    attention_type="default",
    caption_channels=1024,
    cross_attention_dim=768,
    double_self_attention=False,
    dropout=0.0,
    in_channels=4,
    norm_elementwise_affine=False,
    norm_eps=1e-06,
    norm_num_groups=32,
    norm_type="ada_norm_single",
    num_attention_heads=16,
    num_embeds_ada_norm=1000,
    num_layers=12,
    only_cross_attention=False,
    out_channels=8,
    patch_size=2,
    sample_size=128,
    upcast_attention=False,
    use_linear_projection=False
)

save_folder =r'G:\tiny-pixel-art\transformer'
print(model.num_parameters())
model.save_pretrained(save_folder)
model.save_config(save_folder)