Merge branch 'master' into context-windows

comfy/lora.py

@@ -293,6 +293,15 @@ def model_lora_keys_unet(model, key_map={}):
                 key_lora = k[len("diffusion_model."):-len(".weight")]
                 key_map["{}".format(key_lora)] = k
 
+    if isinstance(model, comfy.model_base.QwenImage):
+        for k in sdk:
+            if k.startswith("diffusion_model.") and k.endswith(".weight"): #QwenImage lora format
+                key_lora = k[len("diffusion_model."):-len(".weight")]
+                # Direct mapping for transformer_blocks format (QwenImage LoRA format)
+                key_map["{}".format(key_lora)] = k
+                # Support transformer prefix format
+                key_map["transformer.{}".format(key_lora)] = k
+
     return key_map
 
 

comfy/weight_adapter/lora.py

@@ -96,6 +96,7 @@ class LoRAAdapter(WeightAdapterBase):
         diffusers3_lora = "{}.lora.up.weight".format(x)
         mochi_lora = "{}.lora_B".format(x)
         transformers_lora = "{}.lora_linear_layer.up.weight".format(x)
+        qwen_default_lora = "{}.lora_B.default.weight".format(x)
         A_name = None
 
         if regular_lora in lora.keys():
@@ -122,6 +123,10 @@ class LoRAAdapter(WeightAdapterBase):
             A_name = transformers_lora
             B_name = "{}.lora_linear_layer.down.weight".format(x)
             mid_name = None
+        elif qwen_default_lora in lora.keys():
+            A_name = qwen_default_lora
+            B_name = "{}.lora_A.default.weight".format(x)
+            mid_name = None
 
         if A_name is not None:
             mid = None

comfy_api/latest/_ui.py

@@ -9,7 +9,11 @@ from typing import Type
 import av
 import numpy as np
 import torch
-import torchaudio
+try:
+    import torchaudio
+    TORCH_AUDIO_AVAILABLE = True
+except ImportError:
+    TORCH_AUDIO_AVAILABLE = False
 from PIL import Image as PILImage
 from PIL.PngImagePlugin import PngInfo
 
@@ -302,6 +306,8 @@ class AudioSaveHelper:
 
                 # Resample if necessary
                 if sample_rate != audio["sample_rate"]:
+                    if not TORCH_AUDIO_AVAILABLE:
+                        raise Exception("torchaudio is not available; cannot resample audio.")
                     waveform = torchaudio.functional.resample(waveform, audio["sample_rate"], sample_rate)
 
             # Create output with specified format

comfy_extras/nodes_model_merging_model_specific.py

@@ -314,6 +314,29 @@ class ModelMergeCosmosPredict2_14B(comfy_extras.nodes_model_merging.ModelMergeBl
 
         return {"required": arg_dict}
 
+class ModelMergeQwenImage(comfy_extras.nodes_model_merging.ModelMergeBlocks):
+    CATEGORY = "advanced/model_merging/model_specific"
+
+    @classmethod
+    def INPUT_TYPES(s):
+        arg_dict = { "model1": ("MODEL",),
+                              "model2": ("MODEL",)}
+
+        argument = ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01})
+
+        arg_dict["pos_embeds."] = argument
+        arg_dict["img_in."] = argument
+        arg_dict["txt_norm."] = argument
+        arg_dict["txt_in."] = argument
+        arg_dict["time_text_embed."] = argument
+
+        for i in range(60):
+            arg_dict["transformer_blocks.{}.".format(i)] = argument
+
+        arg_dict["proj_out."] = argument
+
+        return {"required": arg_dict}
+
 NODE_CLASS_MAPPINGS = {
     "ModelMergeSD1": ModelMergeSD1,
     "ModelMergeSD2": ModelMergeSD1, #SD1 and SD2 have the same blocks
@@ -329,4 +352,5 @@ NODE_CLASS_MAPPINGS = {
     "ModelMergeWAN2_1": ModelMergeWAN2_1,
     "ModelMergeCosmosPredict2_2B": ModelMergeCosmosPredict2_2B,
     "ModelMergeCosmosPredict2_14B": ModelMergeCosmosPredict2_14B,
+    "ModelMergeQwenImage": ModelMergeQwenImage,
 }

nodes.py

@@ -1229,12 +1229,12 @@ class RepeatLatentBatch:
         s = samples.copy()
         s_in = samples["samples"]
 
-        s["samples"] = s_in.repeat((amount, 1,1,1))
+        s["samples"] = s_in.repeat((amount,) + ((1,) * (s_in.ndim - 1)))
         if "noise_mask" in samples and samples["noise_mask"].shape[0] > 1:
             masks = samples["noise_mask"]
             if masks.shape[0] < s_in.shape[0]:
-                masks = masks.repeat(math.ceil(s_in.shape[0] / masks.shape[0]), 1, 1, 1)[:s_in.shape[0]]
-            s["noise_mask"] = samples["noise_mask"].repeat((amount, 1,1,1))
+                masks = masks.repeat((math.ceil(s_in.shape[0] / masks.shape[0]),) + ((1,) * (masks.ndim - 1)))[:s_in.shape[0]]
+            s["noise_mask"] = samples["noise_mask"].repeat((amount,) + ((1,) * (samples["noise_mask"].ndim - 1)))
         if "batch_index" in s:
             offset = max(s["batch_index"]) - min(s["batch_index"]) + 1
             s["batch_index"] = s["batch_index"] + [x + (i * offset) for i in range(1, amount) for x in s["batch_index"]]

requirements.txt

@@ -1,5 +1,5 @@
-comfyui-frontend-package==1.23.4
-comfyui-workflow-templates==0.1.51
+comfyui-frontend-package==1.24.4
+comfyui-workflow-templates==0.1.52
 comfyui-embedded-docs==0.2.4
 torch
 torchsde