Update modeling_Llamoe.py

modeling_Llamoe.py

@@ -678,7 +678,7 @@ class LlamoeDecoderLayer(nn.Module):
         super().__init__()
         self.hidden_size = config.hidden_size
 
-        self.self_attn = Llamoe_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+        self.self_attn = LLAMOE_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
 
         self.block_sparse_moe = LlamoeSparseMoeBlock(config)
         self.input_layernorm = LlamoeRMSNorm(config.hidden_size, eps=config.rms_norm_eps)