[Feature] support rl_tp_degree

fastdeploy/model_executor/layers/embeddings.py

@@ -77,6 +77,11 @@ def __init__(
             )
             if self.world_size > 1:
                 set_weight_attrs(self.embeddings.weight, {"output_dim": False})
+            set_weight_attrs(
+                self.embeddings.weight,
+                {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}},
+            )
+
         else:
             # column cut embedding
             self.embeddings = nn.Embedding(

fastdeploy/model_executor/layers/linear.py

@@ -356,11 +356,21 @@ def __init__(
         )
 
         if self.nranks > 0:
+            _set_var_distributed(self.weight, split_axis=-1)
             if self.with_bias:
                 # col parallel
-                _set_var_distributed(self.bias, split_axis=1)
+                _set_var_distributed(self.bias, split_axis=0)
                 set_weight_attrs(self.bias, {"output_dim": True})
 
+        # set_rl_tp_degree
+        set_weight_attrs(
+            self.weight, {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}}
+        )
+        if self.with_bias:
+            set_weight_attrs(
+                self.bias, {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}}
+            )
+
 
 class MergedColumnParallelLinear(ColumnParallelLinear):
     """
@@ -743,6 +753,7 @@ def __init__(
             model_format=fd_config.model_config.model_format,
         )
         if self.nranks > 0:
+            _set_var_distributed(self.weight, split_axis=0)
             if self.with_bias:
                 # col parallel
                 _set_var_distributed(self.bias, split_axis=0)
@@ -755,6 +766,11 @@ def __init__(
 
         self.reduce_results = reduce_results
 
+        # set_rl_tp_degree
+        set_weight_attrs(
+            self.weight, {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}}
+        )
+
     def forward_cuda(self, x: paddle.Tensor) -> paddle.Tensor:
         if self.fd_config.quant_config:
             out = self.quant_method.apply(self, x)

fastdeploy/model_executor/layers/lm_head.py

@@ -94,6 +94,12 @@ def __init__(
                         "model_format": self.fd_config.model_config.model_format,
                     },
                 )
+                if self.bias_key is not None:
+                    set_weight_attrs(
+                        self.linear.bias,
+                        {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}},
+                    )
+
                 if self.nranks > 1:
                     set_weight_attrs(self.linear.weight, {"output_dim": True})
             else:
@@ -116,6 +122,9 @@ def __init__(
 
                 if self.nranks > 1:
                     set_weight_attrs(self.linear.weight, {"output_dim": False})
+        set_weight_attrs(
+            self.linear.weight, {"rl_need_attr": {"rl_tp_degree": fd_config.parallel_config.tensor_parallel_size}}
+        )
 
     def load_state_dict(self, state_dict: Dict[str, paddle.Tensor | np.ndarray]):
         """