[Quant Args] Clean-up

src/compressed_tensors/quantization/lifecycle/initialize.py

@@ -249,15 +249,13 @@ def initialize_qparams(
 
     # 2. Identify quantization scale and zp dtype
     scale_dtype = observed_dtype
-    if quantization_args.scale_dtype is None:
-        if scale_dtype not in [
-            torch.float16,
-            torch.bfloat16,
-            torch.float32,
-            torch.float64,
-        ]:
-            scale_dtype = torch.float16
-        quantization_args.scale_dtype = scale_dtype
+    if scale_dtype not in [
+        torch.float16,
+        torch.bfloat16,
+        torch.float32,
+        torch.float64,
+    ]:
+        scale_dtype = torch.float16
 
     # 3. Initializes scale/zp for the module
     init_scale = Parameter(

src/compressed_tensors/quantization/quant_args.py

@@ -20,7 +20,14 @@
 from compressed_tensors.utils import Aliasable
 from compressed_tensors.utils.helpers import deprecated
 from compressed_tensors.utils.type import TorchDtype
-from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    field_serializer,
+    field_validator,
+    model_validator,
+)
 
 
 __all__ = [
@@ -193,6 +200,12 @@ class QuantizationArgs(BaseModel, use_enum_values=True):
         ),
     )
 
+    @field_serializer("zp_dtype")
+    def serialize_dtype(self, dtype: torch.dtype):
+        if self.symmetric:
+            return None
+        return str(dtype)
+
     @field_validator("type", mode="before")
     def validate_type(cls, value) -> QuantizationType:
         if isinstance(value, str):