Method __init__

architecture/embeddings.py:1571–1581

(
        self,
        image_embed_dim: int = 768,
        cross_attention_dim: int = 768,
        num_image_text_embeds: int = 32,
    )

class ImageProjection(nn.Module):
    def __init__(
        self,
        image_embed_dim: int = 768,
        cross_attention_dim: int = 768,
        num_image_text_embeds: int = 32,
    ):
        super().__init__()

        self.num_image_text_embeds = num_image_text_embeds
        self.image_embeds = nn.Linear(image_embed_dim, self.num_image_text_embeds * cross_attention_dim)
        self.norm = nn.LayerNorm(cross_attention_dim)

    def forward(self, image_embeds: torch.Tensor):
        batch_size = image_embeds.shape[0]