modify feedforward net, move layer norm to start of thr block
This commit is contained in:
parent
e08d8190ae
commit
ddd1b5c0ff
@ -22,7 +22,7 @@ class PyTorchMLPModel(nn.Module):
|
|||||||
def forward(self, x: Tensor) -> Tensor:
|
def forward(self, x: Tensor) -> Tensor:
|
||||||
x = self.relu(self.input_layer(x))
|
x = self.relu(self.input_layer(x))
|
||||||
x = self.dropout(x)
|
x = self.dropout(x)
|
||||||
x = self.relu(self.blocks(x))
|
x = self.blocks(x)
|
||||||
logits = self.output_layer(x)
|
logits = self.output_layer(x)
|
||||||
return logits
|
return logits
|
||||||
|
|
||||||
@ -35,8 +35,8 @@ class Block(nn.Module):
|
|||||||
self.ln = nn.LayerNorm(hidden_dim)
|
self.ln = nn.LayerNorm(hidden_dim)
|
||||||
|
|
||||||
def forward(self, x):
|
def forward(self, x):
|
||||||
x = self.dropout(self.ff(x))
|
x = self.ff(self.ln(x))
|
||||||
x = self.ln(x)
|
x = self.dropout(x)
|
||||||
return x
|
return x
|
||||||
|
|
||||||
|
|
||||||
@ -46,7 +46,6 @@ class FeedForward(nn.Module):
|
|||||||
self.net = nn.Sequential(
|
self.net = nn.Sequential(
|
||||||
nn.Linear(hidden_dim, hidden_dim),
|
nn.Linear(hidden_dim, hidden_dim),
|
||||||
nn.ReLU(),
|
nn.ReLU(),
|
||||||
nn.Linear(hidden_dim, hidden_dim),
|
|
||||||
)
|
)
|
||||||
|
|
||||||
def forward(self, x):
|
def forward(self, x):
|
||||||
|
Loading…
Reference in New Issue
Block a user