nodes/ussoewwin-ComfyUI-QwenImageLoraLoader
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

V3: Always enable AWQ modulation layer LoRA (no switch needed)

Browse Source
This commit is contained in:
ussoewwin
2026-05-17 05:26:52 +09:00
parent 84db2a8abf
commit 7eaa203aab
1 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nodes/lora/qwenimage_v3.py
+6 -1
View File
@@ -207,12 +207,16 @@ class NunchakuQwenImageLoraStackV2:
if model_wrapper.cpu_offload_setting != cpu_offload:
logger.info(f"🔄 Updating CPU offload setting from '{model_wrapper.cpu_offload_setting}' to '{cpu_offload}'")
model_wrapper.cpu_offload_setting = cpu_offload
# V3: Always ensure apply_awq_mod is True (no switch, always enabled)
if hasattr(model_wrapper, "apply_awq_mod") and model_wrapper.apply_awq_mod != True:
logger.info(f"🔄 Updating AWQ mod setting from '{model_wrapper.apply_awq_mod}' to 'True' (V3: always enabled)")
model_wrapper.apply_awq_mod = True
transformer = model_wrapper.model
elif model_wrapper_type_name == "NunchakuQwenImageTransformer2DModel" or model_wrapper_type_name.endswith(
"NunchakuQwenImageTransformer2DModel"
):
logger.info("🔧 Wrapping NunchakuQwenImageTransformer2DModel with ComfyQwenImageWrapper")
logger.info(f"📦 Creating ComfyQwenImageWrapper with cpu_offload='{cpu_offload}'")
logger.info(f"📦 Creating ComfyQwenImageWrapper with cpu_offload='{cpu_offload}', apply_awq_mod=True (V3: always enabled)")
wrapped_model = ComfyQwenImageWrapper(
model_wrapper,
getattr(model_wrapper, "config", {}),
@@ -220,6 +224,7 @@ class NunchakuQwenImageLoraStackV2:
{}, # forward_kwargs
cpu_offload, # cpu_offload_setting
4.0, # vram_margin_gb
apply_awq_mod=True, # V3: Always enable AWQ modulation layer LoRA (no switch needed)
)
model.model.diffusion_model = wrapped_model
model_wrapper = wrapped_model