File size: 15,056 Bytes
4432cc6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 |
import comfy.sd
import comfy.utils
import comfy.model_base
import comfy.model_management
import comfy.model_sampling
import torch
import folder_paths # Assuming this is available from the original context
import json
import os
# It's good practice to check if args is available, especially if running outside full ComfyUI
try:
from comfy.cli_args import args
except ImportError:
class ArgsMock:
disable_metadata = False
args = ArgsMock()
class ModelMergeSimple: # Keeping original for context, not strictly needed for the new node
@classmethod
def INPUT_TYPES(s):
return {"required": { "model1": ("MODEL",),
"model2": ("MODEL",),
"ratio": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),
}}
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, model1, model2, ratio):
m = model1.clone()
kp = model2.get_key_patches("diffusion_model.")
for k in kp:
m.add_patches({k: kp[k]}, ratio, 1.0 - ratio) # Original was 1.0 - ratio, ratio. Swapped to match typical 'ratio applies to model2'
return (m, )
class ModelMergeMultiSimple:
@classmethod
def INPUT_TYPES(s):
inputs = {"required": {}}
for i in range(1, 6): # 5 models
inputs["required"][f"model{i}"] = ("MODEL",)
inputs["required"][f"ratio{i}"] = ("FLOAT", {"default": 0.0, "min": 0.0, "max": 1.0, "step": 0.01})
return inputs
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge_five" # Changed function name to avoid conflict if in same file
CATEGORY = "advanced/model_merging"
def merge_five(self, **kwargs):
models = []
ratios = []
for i in range(1, 6):
model = kwargs.get(f"model{i}")
ratio = kwargs.get(f"ratio{i}")
if model is not None: # Basic check
models.append(model)
ratios.append(ratio)
elif ratio > 0:
# If a ratio is provided for a non-existent model slot (shouldn't happen with type system)
print(f"Warning: Ratio {ratio} provided for model{i} but model is missing. Ignoring.")
ratios.append(0.0) # Effectively ignore
if not models:
raise ValueError("No models provided for merging.")
# Filter out models with a zero ratio to avoid unnecessary processing
# and issues with zero total_ratio if all are zero.
active_models_data = []
for model, ratio in zip(models, ratios):
if ratio > 0:
active_models_data.append({"model": model, "original_ratio": ratio})
if not active_models_data:
print("Warning: All model ratios are 0. Returning the first provided model without changes.")
return (models[0].clone(), )
# Calculate the sum of original ratios for normalization
total_original_ratio = sum(item["original_ratio"] for item in active_models_data)
if total_original_ratio == 0: # Should be caught by previous check, but as a safeguard
print("Warning: Sum of active model ratios is 0. Returning the first provided model.")
return (models[0].clone(), )
# Normalize ratios (these are the w_i in the explanation)
normalized_ratios = [item["original_ratio"] / total_original_ratio for item in active_models_data]
# Start with the first active model
merged_model = active_models_data[0]["model"].clone()
if len(active_models_data) == 1:
# Only one model has a non-zero ratio, just return its clone
return (merged_model,)
current_cumulative_normalized_weight = normalized_ratios[0]
# Iteratively merge subsequent models
for i in range(1, len(active_models_data)):
next_model_data = active_models_data[i]
next_model_normalized_weight = normalized_ratios[i]
# If current_cumulative_normalized_weight is zero (e.g. first model had ratio 0 but others non-zero)
# and this is the first *actual* model to process in the loop.
if current_cumulative_normalized_weight == 0 and i==0 : #This condition might need adjustment based on active_models_data start
merged_model = next_model_data["model"].clone()
current_cumulative_normalized_weight = next_model_normalized_weight
continue # skip the add_patches for this first assignment
# The denominator for scaling factors when adding the next_model
# This is (w_accumulated + w_next)
denominator = current_cumulative_normalized_weight + next_model_normalized_weight
if denominator == 0: # Should not happen if ratios are positive and sum > 0
continue
# Strength for the patches from the next_model (w_next / (w_accumulated + w_next))
strength_for_next_model = next_model_normalized_weight / denominator
# Strength for the patches already in merged_model (w_accumulated / (w_accumulated + w_next))
strength_for_merged_model_self = current_cumulative_normalized_weight / denominator
key_patches = next_model_data["model"].get_key_patches("diffusion_model.")
# ComfyUI's add_patches: m.add_patches({k: kp[k]}, strength_for_incoming_patch, strength_for_self_patch)
# This means: new_value = self_patch * strength_for_self_patch + incoming_patch * strength_for_incoming_patch
for k in key_patches: # Iterate over keys if add_patches doesn't take the whole dict as first arg directly for all keys
merged_model.add_patches({k: key_patches[k]}, strength_for_next_model, strength_for_merged_model_self)
current_cumulative_normalized_weight += next_model_normalized_weight
# Due to potential floating point inaccuracies, it's good to ensure the sum doesn't exceed 1.0
# However, the logic ensures the sum of *normalized original weights* is what we track.
# This cumulative weight is the sum of normalized_ratios of models incorporated so far.
return (merged_model,)
# --- Other classes from your provided code for completeness ---
class ModelSubtract:
@classmethod
def INPUT_TYPES(s):
return {"required": { "model1": ("MODEL",),
"model2": ("MODEL",),
"multiplier": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),
}}
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, model1, model2, multiplier):
m = model1.clone()
kp = model2.get_key_patches("diffusion_model.")
for k in kp:
m.add_patches({k: kp[k]}, multiplier, -multiplier) # Note: ComfyUI add_patches is (patches, mult, mult_self)
# For subtract A - B*mult: m.add_patches({k: kp[k]}, -multiplier, 1.0)
# The example in comfy has it as (A,B,mult) -> A - B*mult + B*mult which is just A
# (A,B,mult) with add_patches(kp, B, A_self) means A_self*m + B*kp
# The ModelSubtract node in Comfy's model_toolרובֿ.py is:
# m.add_patches({k: kp[k]}, -multiplier, 1.0)
# This results in: final_patch = patch_self * 1.0 + patch_other * (-multiplier)
# So, model1_patch - model2_patch * multiplier
# Correcting ModelSubtract based on typical understanding (model1 - model2*multiplier)
# The provided code had: m.add_patches({k: kp[k]}, - multiplier, multiplier)
# This would result in: final = self * mult + other * (-mult)
# Let's assume the user wants model1 - model2*multiplier
m = model1.clone()
kp = model2.get_key_patches("diffusion_model.")
for k in kp:
m.add_patches({k: kp[k]}, -multiplier, 1.0)
return (m, )
class ModelAdd:
@classmethod
def INPUT_TYPES(s):
return {"required": { "model1": ("MODEL",),
"model2": ("MODEL",),
}}
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, model1, model2):
m = model1.clone()
kp = model2.get_key_patches("diffusion_model.")
for k in kp:
m.add_patches({k: kp[k]}, 1.0, 1.0) # model1*1.0 + model2*1.0
return (m, )
# ... (rest of your provided classes: CLIPMergeSimple, CLIPSubtract, CLIPAdd, ModelMergeBlocks, save_checkpoint, CheckpointSave, etc.)
# It's important that these classes are also defined if you are running this as a standalone script for testing,
# or if they are in the same file. For ComfyUI, it will pick them up.
# Make sure to add the new class to NODE_CLASS_MAPPINGS
# Original mappings:
# NODE_CLASS_MAPPINGS = {
# "ModelMergeSimple": ModelMergeSimple,
# "ModelMergeBlocks": ModelMergeBlocks,
# "ModelMergeSubtract": ModelSubtract,
# "ModelMergeAdd": ModelAdd,
# # ... other mappings
# }
# Add the new node:
# Assuming NODE_CLASS_MAPPINGS and NODE_DISPLAY_NAME_MAPPINGS are defined at the end of the file.
# Placeholder for the rest of the code structure
class CLIPMergeSimple:
@classmethod
def INPUT_TYPES(s): return {"required": { "clip1": ("CLIP",),"clip2": ("CLIP",),"ratio": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}), }}
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, clip1, clip2, ratio): return (clip1, ) # Placeholder
class CLIPSubtract:
@classmethod
def INPUT_TYPES(s): return {"required": { "clip1": ("CLIP",),"clip2": ("CLIP",),"multiplier": ("FLOAT", {"default": 1.0, "min": -10.0, "max": 10.0, "step": 0.01}),}}
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, clip1, clip2, multiplier): return (clip1,) # Placeholder
class CLIPAdd:
@classmethod
def INPUT_TYPES(s): return {"required": { "clip1": ("CLIP",),"clip2": ("CLIP",),}}
RETURN_TYPES = ("CLIP",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, clip1, clip2): return (clip1,) # Placeholder
class ModelMergeBlocks:
@classmethod
def INPUT_TYPES(s): return {"required": { "model1": ("MODEL",),"model2": ("MODEL",),"input": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),"middle": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),"out": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01})}}
RETURN_TYPES = ("MODEL",)
FUNCTION = "merge"
CATEGORY = "advanced/model_merging"
def merge(self, model1, model2, **kwargs): return (model1,) # Placeholder
def save_checkpoint(model, clip=None, vae=None, clip_vision=None, filename_prefix=None, output_dir=None, prompt=None, extra_pnginfo=None): pass # Placeholder
class CheckpointSave:
def __init__(self): self.output_dir = folder_paths.get_output_directory()
@classmethod
def INPUT_TYPES(s): return {"required": { "model": ("MODEL",),"clip": ("CLIP",),"vae": ("VAE",),"filename_prefix": ("STRING", {"default": "checkpoints/ComfyUI"}),},"hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},}
RETURN_TYPES = ()
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
def save(self, model, clip, vae, filename_prefix, prompt=None, extra_pnginfo=None): return {} # Placeholder
class CLIPSave:
def __init__(self): self.output_dir = folder_paths.get_output_directory()
@classmethod
def INPUT_TYPES(s): return {"required": { "clip": ("CLIP",),"filename_prefix": ("STRING", {"default": "clip/ComfyUI"}),},"hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},}
RETURN_TYPES = ()
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
def save(self, clip, filename_prefix, prompt=None, extra_pnginfo=None): return {} # Placeholder
class VAESave:
def __init__(self): self.output_dir = folder_paths.get_output_directory()
@classmethod
def INPUT_TYPES(s): return {"required": { "vae": ("VAE",),"filename_prefix": ("STRING", {"default": "vae/ComfyUI_vae"}),},"hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},}
RETURN_TYPES = ()
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
def save(self, vae, filename_prefix, prompt=None, extra_pnginfo=None): return {} # Placeholder
class ModelSave:
def __init__(self): self.output_dir = folder_paths.get_output_directory()
@classmethod
def INPUT_TYPES(s): return {"required": { "model": ("MODEL",),"filename_prefix": ("STRING", {"default": "diffusion_models/ComfyUI"}),},"hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},}
RETURN_TYPES = ()
FUNCTION = "save"
OUTPUT_NODE = True
CATEGORY = "advanced/model_merging"
def save(self, model, filename_prefix, prompt=None, extra_pnginfo=None): return {} # Placeholder
NODE_CLASS_MAPPINGS = {
"ModelMergeSimple": ModelMergeSimple,
"ModelMergeMultiSimple": ModelMergeMultiSimple, # Added new class
"ModelMergeBlocks": ModelMergeBlocks,
"ModelMergeSubtract": ModelSubtract,
"ModelMergeAdd": ModelAdd,
"CheckpointSave": CheckpointSave,
"CLIPMergeSimple": CLIPMergeSimple,
"CLIPMergeSubtract": CLIPSubtract,
"CLIPMergeAdd": CLIPAdd,
"CLIPSave": CLIPSave,
"VAESave": VAESave,
"ModelSave": ModelSave,
}
NODE_DISPLAY_NAME_MAPPINGS = {
"ModelMergeSimple": "Model Merge Simple (2 Models)", # Clarified original
"ModelMergeMultiSimple": "Model Merge Multi Simple (5 Models)", # Added new display name
"ModelMergeBlocks": "Model Merge Blocks",
"ModelMergeSubtract": "Model Subtract",
"ModelMergeAdd": "Model Add",
"CheckpointSave": "Save Checkpoint",
"CLIPMergeSimple": "CLIP Merge Simple",
"CLIPMergeSubtract": "CLIP Subtract",
"CLIPMergeAdd": "CLIP Add",
"CLIPSave": "CLIP Save",
"VAESave": "VAE Save",
"ModelSave": "Model Save",
}
print("Custom model merging nodes loaded.") |