Enforce that TP > 1 is not supported for Mamba2 if Quantization is Enabled. (#14617)

Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>

vllm/model_executor/layers/mamba/mamba_mixer2.py

@@ -251,6 +251,9 @@ class MambaMixer2(CustomOp):
                 "then num_groups must equal 1."
             )
 
+        assert self.tp_size == 1 or quant_config is None, \
+            "Tensor parallel currently not supported for quantized models."
+
         self.ssm_state_size = ssm_state_size
         self.activation = activation
 
@@ -331,6 +334,8 @@ class MambaMixer2(CustomOp):
                 ], self.tp_size, tp_rank)
             })
 
+        if quant_config is None:
+            # - quant layers do not have a weight loader
             delattr(self.in_proj.weight, "weight_loader")
             set_weight_attrs(
                 self.in_proj.weight,