Update evo_model.py

evo_model.py

@@ -6,23 +6,25 @@ class EvoEncoder(nn.Module):
     def __init__(self, d_model=512, num_heads=8, ffn_dim=1024, num_layers=6, memory_enabled=True):
         super().__init__()
         self.embedding = nn.Embedding(30522, d_model)
-        self.positional_encoding = nn.Parameter(torch.zeros(1, 512, d_model))
-        encoder_layer = nn.TransformerEncoderLayer(d_model=d_model, nhead=num_heads, dim_feedforward=ffn_dim, batch_first=True)
+        encoder_layer = nn.TransformerEncoderLayer(
+            d_model=d_model,
+            nhead=num_heads,
+            dim_feedforward=ffn_dim,
+            batch_first=True
+        )
         self.transformer = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
-        self.norm = nn.LayerNorm(d_model)
         self.memory_enabled = memory_enabled
         if memory_enabled:
             self.memory_proj = nn.Linear(d_model, d_model)
             self.memory_token = nn.Parameter(torch.zeros(1, 1, d_model))
 
     def forward(self, input_ids):
-        x = self.embedding(input_ids) + self.positional_encoding[:, :input_ids.size(1), :]
+        x = self.embedding(input_ids)
         if self.memory_enabled:
             mem = self.memory_token.expand(x.size(0), -1, -1)
             x = torch.cat([mem, x], dim=1)
         x = self.transformer(x)
-        x = self.norm(x)
-        return x[:, 0]  # Return [CLS]-like token (memory or first token)
+        return x[:, 0]  # Return memory token or first token
 
 class EvoTransformer(nn.Module):
     def __init__(self, d_model=512, num_heads=8, ffn_dim=1024, num_layers=6, num_classes=1, memory_enabled=True):