Can I change the T5 embedding class like this code where they have changed the BERT embedding class(visual+text embedding)?

Question

In this code they have concatenated the visual embedding with the text embedding

vis_embeddings = self.visualLayer(
        visual_embd.view(-1, self.config.monet_hidden_size)
    ).view(input_shape[0], 200, self.config.hidden_size)
    assert vis_embeddings.shape[1] == visual_embd.shape[1]

    # vis_embeddings = vis_embeddings + self.frame_embeddings(torch.tensor(self.frames).unsqueeze(0).to(inputs_embeds.device))

    # concat
    embeddings = torch.cat((inputs_embeds, vis_embeddings), dim=1)
    embeddings = embeddings + self.token_type_embeddings(token_type_ids)

    embeddings = torch.cat((cls_embedding, embeddings), dim=1)

    if self.position_embedding_type == "absolute":
        # print('error')
        position_embeddings = self.position_embeddings(position_ids)
        # print(position_embeddings.shape, embeddings.shape)
        # assert position_embeddings.shape==embeddings.shape
        embeddings += position_embeddings
    embeddings = self.LayerNorm(embeddings)
    embeddings = self.dropout(embeddings)
    return embeddings

Can this code be replicated in t5?