Revert "fix: adjust memory_usage_factor for quantized models"

This reverts commit a07f0fb.

Author: silveroxides

nodes/loader_nodes.py

@@ -124,21 +124,6 @@ def load_checkpoint(
         clip = out[1]
         vae = out[2]
 
-        # Adjust memory_usage_factor for quantized models
-        # Quantized weights use ~1 byte (INT8/FP8) vs 2 bytes (bf16/fp16)
-        # This prevents ComfyUI from over-estimating memory requirements
-        if model is not None and quant_format != "auto":
-            try:
-                original_factor = model.model.memory_usage_factor
-                # INT8/FP8 weights are ~2x smaller than fp16/bf16
-                model.model.memory_usage_factor = original_factor / 2.0
-                logging.info(
-                    f"QuantizedModelLoader: Adjusted memory_usage_factor from "
-                    f"{original_factor:.2f} to {model.model.memory_usage_factor:.2f}"
-                )
-            except AttributeError:
-                logging.debug("Could not adjust memory_usage_factor")
-
         # Force dequantize if requested (useful for debugging)
         if force_dequant and model is not None:
             logging.info("QuantizedModelLoader: Force dequantizing model weights")
@@ -231,19 +216,6 @@ def load_unet(self, unet_name, quant_format, kernel_backend):
         # Standard loading path
         model = comfy.sd.load_diffusion_model(unet_path, model_options=model_options)
 
-        # Adjust memory_usage_factor for quantized models
-        # Quantized weights use ~1 byte (INT8/FP8) vs 2 bytes (bf16/fp16)
-        if model is not None and quant_format != "auto":
-            try:
-                original_factor = model.model.memory_usage_factor
-                model.model.memory_usage_factor = original_factor / 2.0
-                logging.info(
-                    f"QuantizedUNETLoader: Adjusted memory_usage_factor from "
-                    f"{original_factor:.2f} to {model.model.memory_usage_factor:.2f}"
-                )
-            except AttributeError:
-                logging.debug("Could not adjust memory_usage_factor")
-
         return (model,)
 
 

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "ComfyUI-QuantOps"
 description = "Extended quantization layouts for ComfyUI (INT8, row/block-wise FP8)"
-version = "1.0.4"
+version = "1.0.3"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.9"