关于Z-Image Omni代码中的文本编码长度问题

def _pad_with_ids(
        self,
        feat: torch.Tensor,
        pos_grid_size: Tuple,
        pos_start: Tuple,
        device: torch.device,
        noise_mask_val: Optional[int] = None,
    ):
        """Pad feature to SEQ_MULTI_OF, create position IDs and pad mask."""
        ori_len = len(feat)
        pad_len = (-ori_len) % SEQ_MULTI_OF
        total_len = ori_len + pad_len

# Process captions
for j, cap_item in enumerate(all_cap_feats[i]):
    noise_val = images_noise_mask[i][j] if j < len(images_noise_mask[i]) else 1
    cap_out, cap_pos, cap_mask, cap_len, cap_nm = self._pad_with_ids(
        cap_item,
        (len(cap_item) + (-len(cap_item)) % SEQ_MULTI_OF, 1, 1),
        (cap_cu_len, 0, 0),
        device,
        noise_val,
    )
    cap_feats_list.append(cap_out)
    cap_pos_list.append(cap_pos)

我在研究Omni源码时候发现，对于文本编码的特征。_pad_with_ids函数入参的时候就已经对齐了SEQ_MULTI_OF=32的倍数，但是进入之后还是会pad_len = (-ori_len) % SEQ_MULTI_OF继续对齐，导致输出cap_out和cap_pos的长度并不一样，这是否有问题。同时我看图像的vae和siglip特征并没有这个问题

Start at _pad_with_ids and the caption-processing loop shown in the issue. Trace the lengths passed for pre-aligned text features and compare cap_out with cap_pos, then inspect the corresponding VAE and SigLIP paths. Done means confirming whether the mismatch is real and documenting or correcting it with a focused regression check.