diff --git a/fastdeploy/input/text_processor.py b/fastdeploy/input/text_processor.py
index 97aac5cf6f2..a1baf8e4658 100644
--- a/fastdeploy/input/text_processor.py
+++ b/fastdeploy/input/text_processor.py
@@ -185,6 +185,9 @@ def __init__(self, model_name_or_path, reasoning_parser_obj=None, tool_parser_ob
         from paddleformers.trl.llm_utils import get_eos_token_id
 
         self.eos_token_ids = get_eos_token_id(self.tokenizer, self.generation_config)
+        data_processor_logger.info(
+            f"The eos_token_ids obtained by merging tokenizer and generation_config is {self.eos_token_ids}"
+        )
         self.eos_token_id_len = len(self.eos_token_ids)
         self.pad_token_id = self.get_pad_id()
         self.reasoning_parser = None
@@ -396,7 +399,7 @@ def process_response_dict_normal(self, response_dict, **kwargs):
         is_end = response_dict["finished"]
         req_id = response_dict["request_id"]
         if is_end and len(token_ids) > 0 and not kwargs.get("include_stop_str_in_output"):
-            if token_ids[-1] == self.tokenizer.eos_token_id:
+            if token_ids[-1] in self.eos_token_ids:
                 token_ids = token_ids[:-1]
         delta_text, _, previous_texts = self.ids2tokens(token_ids, req_id)
         if is_end:
@@ -434,7 +437,7 @@ def process_response_dict_streaming(self, response_dict, **kwargs):
         token_ids = response_dict["outputs"]["token_ids"]
 
         if is_end and len(token_ids) > 0 and not kwargs.get("include_stop_str_in_output"):
-            if token_ids[-1] == self.tokenizer.eos_token_id:
+            if token_ids[-1] in self.eos_token_ids:
                 token_ids = token_ids[:-1]
         delta_text, previous_token_ids, previous_texts = self.ids2tokens(token_ids, req_id)
         response_dict["outputs"]["raw_prediction"] = delta_text
diff --git a/fastdeploy/model_executor/layers/moe/fused_moe_backend_base.py b/fastdeploy/model_executor/layers/moe/fused_moe_backend_base.py
index ea97bb5d739..d1be7af8036 100644
--- a/fastdeploy/model_executor/layers/moe/fused_moe_backend_base.py
+++ b/fastdeploy/model_executor/layers/moe/fused_moe_backend_base.py
@@ -199,6 +199,7 @@ def create_weights(self, layer: nn.Layer, **extra_weight_attrs):
             layer.up_gate_proj_weight,
             {
                 "weight_loader": extra_weight_attrs.get("weight_loader", default_weight_loader(layer.fd_config)),
+                "weight_need_transpose": extra_weight_attrs.get("model_format") == "torch",
                 "model_format": extra_weight_attrs.get("model_format", ""),
             },
         )
@@ -206,6 +207,7 @@ def create_weights(self, layer: nn.Layer, **extra_weight_attrs):
             layer.down_proj_weight,
             {
                 "weight_loader": extra_weight_attrs.get("weight_loader", default_weight_loader(layer.fd_config)),
+                "weight_need_transpose": extra_weight_attrs.get("model_format") == "torch",
                 "model_format": extra_weight_attrs.get("model_format", ""),
             },
         )
diff --git a/fastdeploy/model_executor/layers/quantization/__init__.py b/fastdeploy/model_executor/layers/quantization/__init__.py
index 6be17828290..f8716369852 100644
--- a/fastdeploy/model_executor/layers/quantization/__init__.py
+++ b/fastdeploy/model_executor/layers/quantization/__init__.py
@@ -85,6 +85,8 @@ def parse_quant_config(args, model_config, is_ernie, is_v1_loader):
     else:
         if not quantization_config.get("is_quantized"):
             quantization_config["is_quantized"] = model_config.is_quantized
+        if args.dynamic_load_weight and quantization_config is not None:
+            quantization_config["is_quantized"] = True
         quant_cls = get_quantization_config(quant_config_name)
         quant_config = quant_cls.from_config(quantization_config)
     return quant_config