Update modeling_gemmoe.py

modeling_gemmoe.py

@@ -743,22 +743,12 @@ class GemmoeDecoderLayer(nn.Module):
             output_attentions=output_attentions,
             use_cache=use_cache,
         )
-        
-        # Check if the tensor sizes match before adding residual
-        if hidden_states.size() != residual.size():
-            hidden_states = hidden_states[:, -residual.size(1):, :]
-        
         hidden_states = residual + hidden_states
 
         # Fully Connected
         residual = hidden_states
         hidden_states = self.post_attention_layernorm(hidden_states)
         hidden_states, router_logits = self.block_sparse_moe(hidden_states)
-        
-        # Check if the tensor sizes match before adding residual
-        if hidden_states.size() != residual.size():
-            hidden_states = hidden_states[:, -residual.size(1):, :]
-        
         hidden_states = residual + hidden_states
 
         outputs = (hidden_states,)
@@ -775,8 +765,6 @@ class GemmoeDecoderLayer(nn.Module):
         return outputs
 
 
-
-
 GEMMOE_START_DOCSTRING = r"""
     This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
     library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads