sat 版本是否不支持 i2v?

As the title suggests, in sat/sample_video.py, model_config.network_config.params.in_channels is hard-coded to 16 (configs/cogvideox_2b.yaml, lines 29-34), and there is no extra channel reserved. Although DynamicCFG does support cond["concat"], once we fill in the image latent, we will encounter the following error: expected input[...] to have 16 channels, but got 32, because the model weights do not have corresponding extra convolution kernels. Therefore, the SAT version currently does not have a true i2v branch. Although the code encodes the image into latent (concat_images), it has not been merged with the noise latent, which is equivalent to pure t2v. Is that correct?