Move create_parameters to __init__ in FuseMOE for CultassBackend and TritonBackend (#3148)

* w4a8 bug

* fix w4a8 bug

* remove code

* modify the triton backend

* fix ep

* fix the bug with tensor_wise_fp8 in triton backend

* fix the RL

* fix bug by merge

* fix the bug in w4a8

* fix the tensor_wise_fp8 bug

* fix RL

fastdeploy/model_executor/layers/utils.py

@@ -15,7 +15,7 @@
 """
 
 import functools
-from typing import Tuple, Union
+from typing import Any, Optional, Tuple, Union
 
 import numpy as np
 import paddle
@@ -45,6 +45,14 @@ if cache_params != "none":
     c8_state_dict = paddle.load(cache_params, return_numpy=True)
 
 
+# TODO(lulinjun): delete it, import from fastdeploy.model_executor.models.utils after supporting all backends
+def set_weight_attrs(param, param_attr_map: Optional[dict[str, Any]]):
+    if param_attr_map is None:
+        return
+    for key, value in param_attr_map.items():
+        setattr(param, key, value)
+
+
 def per_block_cast_to_fp8(x: Tensor, block_size: list = [128, 128]) -> Tuple[Tensor, Tensor]:
     """
     Only used in deep_gemm block wise quant weight.