Method update

python/dgl/optim/pytorch/sparse_optim.py:436–449

Update embeddings in a sparse manner Sparse embeddings are updated in mini batches. We maintain gradient states for each embedding so they can be updated separately. Parameters ---------- idx : tensor Index of the embeddings to be updated.

(self, idx, grad, emb)

    @abstractmethod
    def update(self, idx, grad, emb):
        """Update embeddings in a sparse manner
        Sparse embeddings are updated in mini batches. We maintain gradient states for
        each embedding so they can be updated separately.

        Parameters
        ----------
        idx : tensor
            Index of the embeddings to be updated.
        grad : tensor
            Gradient of each embedding.
        emb : dgl.nn.NodeEmbedding
            Sparse node embedding to update.
        """

    def zero_grad(self):
        """clean grad cache"""