(self, x: torch.Tensor)
| 95 | self.w3 = nn.Linear(dim, hidden_dim, bias=False) # up |
| 96 | |
| 97 | def forward(self, x: torch.Tensor) -> torch.Tensor: |
| 98 | return self.w2(F.silu(self.w1(x)) * self.w3(x)) |
| 99 | |
| 100 | |
| 101 | class TransformerBlock(nn.Module): |
nothing calls this directly
no outgoing calls
no test coverage detected