Update utils.py

Co-authored-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>

Author: LucasWilkinson

vllm/attention/backends/mla/utils.py

@@ -231,15 +231,15 @@ def process_weights_after_loading(self):
                 .flatten(start_dim=0, end_dim=1).contiguous()
 
             tp_size = get_tensor_model_parallel_world_size()
-            self.o_proj_absored = RowParallelLinear(
+            self.o_proj_absorbed = RowParallelLinear(
                 self.W_UV_O.shape[0] * tp_size,
                 self.W_UV_O.shape[1],
                 bias=False,
                 # TODO(lucas) figure out how to properly forward quant_method
                 #quant_config=self.o_proj.quant_method,
             )
 
-            self.o_proj_absored.weight = torch.nn.Parameter(self.W_UV_O.T)
+            self.o_proj_absorbed.weight = torch.nn.Parameter(self.W_UV_O.T)
         else:
             self.W_UV = W_UV
             self.W_UK = W_UK