Method forward

inference/models/mpt.py:89–92

(self, input: torch.Tensor, unembed: bool = False)

class SharedEmbedding(nn.Embedding):
    def forward(self, input: torch.Tensor, unembed: bool = False) -> torch.Tensor:
        if unembed:
            return F.linear(input, self.weight)
        return super().forward(input)


class MPTAttentionFused(nn.Module):