Schematic of two tower user and item encoder

gistfile1.txt

    # Compute the user embedding
    user_embedding = self.compute_user_embedding(
        user_id, user_features, user_history
    )  # [B, DU]
    # Compute item embeddings
    item_embeddings = self.compute_item_embeddings(
        item_id, item_features
    )  # [B, DI]
    # Compute the scores for every pair of user and item
    scores = torch.matmul(user_embedding, item_embeddings.t())