[V1 Loader] Support loading static C8 scale JSON (#5909)

* v1 loader: support loading static C8 scale JSON

* update

fastdeploy/model_executor/load_weight_utils.py

@@ -76,7 +76,12 @@ def get_weight_iterator(model_path: str):
         weights_iterator = safetensors_weights_iterator(files_list)
     else:
         weights_iterator = pdparams_weight_iterator(files_list)
-    return weights_iterator
+
+    yield from weights_iterator
+
+    kv_cache_scale_json_path = Path(model_path) / "kv_cache_scale.json"
+    if kv_cache_scale_json_path.exists():
+        yield from kv_cache_scale_iterator(str(kv_cache_scale_json_path))
 
 
 def is_weight_cache_enabled(fd_config, weight_cache_path=".cache"):
@@ -319,6 +324,17 @@ def load_ep_checkpoint(cls: PretrainedModel, model_path: str, fd_config: FDConfi
     return state_dict
 
 
+def kv_cache_scale_iterator(kv_cache_scale_json_path):
+    """
+    kv_cache_scale_iterator
+    """
+    with open(kv_cache_scale_json_path, "r") as f:
+        data = json.load(f)
+        for key, value in data.items():
+            scale_tensor = paddle.to_tensor(value, dtype=paddle.get_default_dtype()) * 448.0
+            yield key, scale_tensor
+
+
 def safetensors_weights_iterator(safe_tensor_list: list[str]):
     """
     safetensors_weights_iterator
@@ -399,7 +415,7 @@ def deal_state_dict(state_dict):
             src_tensor._share_data_with(dst_tensor)
 
 
-def load_cache_scale(fd_config, state_dict):
+def load_kv_cache_scale(fd_config, state_dict):
     file_path = fd_config.model_config.kv_cache_quant_scale_path
     prefix_layer_name = fd_config.model_config.prefix_layer_name
     if os.path.exists(file_path):
@@ -465,6 +481,6 @@ def load_composite_checkpoint(
     if hasattr(fd_config.quant_config, "kv_cache_quant_type"):
         kv_cache_quant_type = fd_config.quant_config.kv_cache_quant_type
         if kv_cache_quant_type == "float8_e4m3fn":
-            load_cache_scale(fd_config, state_dict)
+            load_kv_cache_scale(fd_config, state_dict)
 
     return state_dict