def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
if self.training:
m = x.bernoulli(self._word_dropout_ratio)
x = torch.where(m == 1, torch.tensor(0).to(x.device), x)
static = self._static(x).permute(0, 2, 1)
non_static = self._non_static(x).permute(0, 2, 1)
return static, non_static
def forward(self, x: torch.Tensor) -> torch.Tensor:
if self.training:
m = x.bernoulli(self._word_dropout_ratio)
x = torch.where(m == 1, torch.tensor(0).to(x.device), x)
embedding = self._embedding(x).permute(0, 2, 1)
r = self._conv(embedding)
r = F.relu(r)
r = self._maxpool(r)
r = self._conv1(r)
r = F.relu(r)
r = self._maxpool1(r)
r = r.permute(0, 2, 1)
_, r = self._bilstm(r)
feature = torch.cat([*r[0]], dim=1)
r = self._dropout(feature)
score = self._fc(r)
return score
