Add Lora Loader node support for Nunchaku SVDQuant FLUX model architecture with template workflow. Fixes #255

py/nodes/lora_loader.py

@@ -2,14 +2,14 @@ import logging
 from nodes import LoraLoader
 from comfy.comfy_types import IO # type: ignore
 import asyncio
-from .utils import FlexibleOptionalInputType, any_type, get_lora_info, extract_lora_name, get_loras_list
+from .utils import FlexibleOptionalInputType, any_type, get_lora_info, extract_lora_name, get_loras_list, nunchaku_load_lora
 
 logger = logging.getLogger(__name__)
 
 class LoraManagerLoader:
     NAME = "Lora Loader (LoraManager)"
     CATEGORY = "Lora Manager/loaders"
-
+    
     @classmethod
     def INPUT_TYPES(cls):
         return {
@@ -37,19 +37,39 @@ class LoraManagerLoader:
         
         clip = kwargs.get('clip', None)
         lora_stack = kwargs.get('lora_stack', None)
+        
+        # Check if model is a Nunchaku Flux model - simplified approach
+        is_nunchaku_model = False
+        
+        try:
+            model_wrapper = model.model.diffusion_model
+            # Check if model is a Nunchaku Flux model using only class name
+            if model_wrapper.__class__.__name__ == "ComfyFluxWrapper":
+                is_nunchaku_model = True
+                logger.info("Detected Nunchaku Flux model")
+        except (AttributeError, TypeError):
+            # Not a model with the expected structure
+            pass
+        
         # First process lora_stack if available
         if lora_stack:
             for lora_path, model_strength, clip_strength in lora_stack:
-                # Apply the LoRA using the provided path and strengths
-                model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+                # Apply the LoRA using the appropriate loader
+                if is_nunchaku_model:
+                    # Use our custom function for Flux models
+                    model = nunchaku_load_lora(model, lora_path, model_strength)
+                    # clip remains unchanged for Nunchaku models
+                else:
+                    # Use default loader for standard models
+                    model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
                 
                 # Extract lora name for trigger words lookup
                 lora_name = extract_lora_name(lora_path)
                 _, trigger_words = asyncio.run(get_lora_info(lora_name))
                 
                 all_trigger_words.extend(trigger_words)
-                # Add clip strength to output if different from model strength
-                if abs(model_strength - clip_strength) > 0.001:
+                # Add clip strength to output if different from model strength (except for Nunchaku models)
+                if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
                     loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
                 else:
                     loaded_loras.append(f"{lora_name}: {model_strength}")
@@ -68,11 +88,17 @@ class LoraManagerLoader:
             # Get lora path and trigger words
             lora_path, trigger_words = asyncio.run(get_lora_info(lora_name))
             
-            # Apply the LoRA using the resolved path with separate strengths
-            model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
+            # Apply the LoRA using the appropriate loader
+            if is_nunchaku_model:
+                # For Nunchaku models, use our custom function
+                model = nunchaku_load_lora(model, lora_path, model_strength)
+                # clip remains unchanged
+            else:
+                # Use default loader for standard models
+                model, clip = LoraLoader().load_lora(model, clip, lora_path, model_strength, clip_strength)
             
-            # Include clip strength in output if different from model strength
-            if abs(model_strength - clip_strength) > 0.001:
+            # Include clip strength in output if different from model strength and not a Nunchaku model
+            if not is_nunchaku_model and abs(model_strength - clip_strength) > 0.001:
                 loaded_loras.append(f"{lora_name}: {model_strength},{clip_strength}")
             else:
                 loaded_loras.append(f"{lora_name}: {model_strength}")

py/nodes/utils.py

@@ -35,7 +35,12 @@ any_type = AnyType("*")
 # Common methods extracted from lora_loader.py and lora_stacker.py
 import os
 import logging
-import asyncio
+import copy
+import folder_paths
+import torch
+import safetensors.torch
+from diffusers.utils.state_dict_utils import convert_unet_state_dict_to_peft
+from diffusers.loaders import FluxLoraLoaderMixin
 from ..services.lora_scanner import LoraScanner
 from ..config import config
 
@@ -81,4 +86,64 @@ def get_loras_list(kwargs):
     # Unexpected format
     else:
         logger.warning(f"Unexpected loras format: {type(loras_data)}")
-        return []
+        return []
+
+def load_state_dict_in_safetensors(path, device="cpu", filter_prefix=""):
+    """Simplified version of load_state_dict_in_safetensors that just loads from a local path"""  
+    state_dict = {}
+    with safetensors.torch.safe_open(path, framework="pt", device=device) as f:
+        for k in f.keys():
+            if filter_prefix and not k.startswith(filter_prefix):
+                continue
+            state_dict[k.removeprefix(filter_prefix)] = f.get_tensor(k)
+    return state_dict
+
+def to_diffusers(input_lora):
+    """Simplified version of to_diffusers for Flux LoRA conversion"""
+    if isinstance(input_lora, str):
+        tensors = load_state_dict_in_safetensors(input_lora, device="cpu")
+    else:
+        tensors = {k: v for k, v in input_lora.items()}
+
+    # Convert FP8 tensors to BF16
+    for k, v in tensors.items():
+        if v.dtype not in [torch.float64, torch.float32, torch.bfloat16, torch.float16]:
+            tensors[k] = v.to(torch.bfloat16)
+    
+    new_tensors = FluxLoraLoaderMixin.lora_state_dict(tensors)
+    new_tensors = convert_unet_state_dict_to_peft(new_tensors)
+
+    return new_tensors
+
+def nunchaku_load_lora(model, lora_name, lora_strength):
+    """Load a Flux LoRA for Nunchaku model"""   
+    model_wrapper = model.model.diffusion_model
+    transformer = model_wrapper.model
+    
+    # Save the transformer temporarily
+    model_wrapper.model = None
+    ret_model = copy.deepcopy(model)  # copy everything except the model
+    ret_model_wrapper = ret_model.model.diffusion_model
+    
+    # Restore the model and set it for the copy
+    model_wrapper.model = transformer
+    ret_model_wrapper.model = transformer
+    
+    # Get full path to the LoRA file
+    lora_path = folder_paths.get_full_path("loras", lora_name)
+    ret_model_wrapper.loras.append((lora_path, lora_strength))
+    
+    # Convert the LoRA to diffusers format
+    sd = to_diffusers(lora_path)
+    
+    # Handle embedding adjustment if needed
+    if "transformer.x_embedder.lora_A.weight" in sd:
+        new_in_channels = sd["transformer.x_embedder.lora_A.weight"].shape[1]
+        assert new_in_channels % 4 == 0
+        new_in_channels = new_in_channels // 4
+        
+        old_in_channels = ret_model.model.model_config.unet_config["in_channels"]
+        if old_in_channels < new_in_channels:
+            ret_model.model.model_config.unet_config["in_channels"] = new_in_channels
+    
+    return ret_model

pyproject.toml

@@ -7,6 +7,7 @@ dependencies = [
     "aiohttp",
     "jinja2", 
     "safetensors",
+    "diffusers",
     "watchdog",
     "beautifulsoup4",
     "piexif",