Address some comments

Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>

Author: shanjiaz

src/compressed_tensors/compressors/quantized_compressors/base.py

@@ -131,7 +131,8 @@ def compress(
                 if name in compressed_param_names:
                     continue
 
-                # omit saving zero points for symmetric quantization
+                # for symmetric quantization, omit zero_point
+                # manually because it wasn't handled in compress_weight
                 if name.endswith("weight_zero_point"):
                     module_path = name.rsplit(".", 1)[0]
                     if (

src/compressed_tensors/compressors/quantized_compressors/fp4_quantized.py

@@ -118,7 +118,6 @@ def compress_weight(
                 "NVFP4 quantization requires global_scale (TENSOR_GROUP strategy). "
                 "Use TENSOR_GROUP strategy instead of GROUP for FP4 quantization."
             )
-        compressed_dict["weight_global_scale"] = global_scale
 
         return compressed_dict