Merge upstream/master, keep local README.md

.ci/update_windows/update.py

@@ -53,6 +53,16 @@ try:
     repo.stash(ident)
 except KeyError:
     print("nothing to stash")  # noqa: T201
+except:
+    print("Could not stash, cleaning index and trying again.")  # noqa: T201
+    repo.state_cleanup()
+    repo.index.read_tree(repo.head.peel().tree)
+    repo.index.write()
+    try:
+        repo.stash(ident)
+    except KeyError:
+        print("nothing to stash.")  # noqa: T201
+
 backup_branch_name = 'backup_branch_{}'.format(datetime.today().strftime('%Y-%m-%d_%H_%M_%S'))
 print("creating backup branch: {}".format(backup_branch_name))  # noqa: T201
 try:

api_server/routes/internal/internal_routes.py

@@ -58,8 +58,13 @@ class InternalRoutes:
                 return web.json_response({"error": "Invalid directory type"}, status=400)
 
             directory = get_directory_by_type(directory_type)
+
+            def is_visible_file(entry: os.DirEntry) -> bool:
+                """Filter out hidden files (e.g., .DS_Store on macOS)."""
+                return entry.is_file() and not entry.name.startswith('.')
+
             sorted_files = sorted(
-                (entry for entry in os.scandir(directory) if entry.is_file()),
+                (entry for entry in os.scandir(directory) if is_visible_file(entry)),
                 key=lambda entry: -entry.stat().st_mtime
             )
             return web.json_response([entry.name for entry in sorted_files], status=200)

comfy/model_detection.py

@@ -259,8 +259,10 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
                 dit_config["nerf_tile_size"] = 512
                 dit_config["nerf_final_head_type"] = "conv" if f"{key_prefix}nerf_final_layer_conv.norm.scale" in state_dict_keys else "linear"
                 dit_config["nerf_embedder_dtype"] = torch.float32
-            if "__x0__" in state_dict_keys: # x0 pred
+                if "__x0__" in state_dict_keys: # x0 pred
-                dit_config["use_x0"] = True
+                    dit_config["use_x0"] = True
+                else:
+                    dit_config["use_x0"] = False
         else:
             dit_config["guidance_embed"] = "{}guidance_in.in_layer.weight".format(key_prefix) in state_dict_keys
             dit_config["yak_mlp"] = '{}double_blocks.0.img_mlp.gate_proj.weight'.format(key_prefix) in state_dict_keys

comfy/sd.py

@@ -549,8 +549,10 @@ class VAE:
                     ddconfig = {"dim": dim, "z_dim": self.latent_channels, "dim_mult": [1, 2, 4, 4], "num_res_blocks": 2, "attn_scales": [], "temperal_downsample": [False, True, True], "dropout": 0.0}
                     self.first_stage_model = comfy.ldm.wan.vae.WanVAE(**ddconfig)
                     self.working_dtypes = [torch.bfloat16, torch.float16, torch.float32]
-                    self.memory_used_encode = lambda shape, dtype: 6000 * shape[3] * shape[4] * model_management.dtype_size(dtype)
+                    self.memory_used_encode = lambda shape, dtype: (1500 if shape[2]<=4 else 6000) * shape[3] * shape[4] * model_management.dtype_size(dtype)
-                    self.memory_used_decode = lambda shape, dtype: 7000 * shape[3] * shape[4] * (8 * 8) * model_management.dtype_size(dtype)
+                    self.memory_used_decode = lambda shape, dtype: (2200 if shape[2]<=4 else 7000) * shape[3] * shape[4] * (8*8) * model_management.dtype_size(dtype)
+
+
             # Hunyuan 3d v2 2.0 & 2.1
             elif "geo_decoder.cross_attn_decoder.ln_1.bias" in sd:
 

comfy/supported_models.py

@@ -541,7 +541,7 @@ class SD3(supported_models_base.BASE):
     unet_extra_config = {}
     latent_format = latent_formats.SD3
 
-    memory_usage_factor = 1.2
+    memory_usage_factor = 1.6
 
     text_encoder_key_prefix = ["text_encoders."]
 
@@ -965,7 +965,7 @@ class CosmosT2IPredict2(supported_models_base.BASE):
 
     def __init__(self, unet_config):
         super().__init__(unet_config)
-        self.memory_usage_factor = (unet_config.get("model_channels", 2048) / 2048) * 0.9
+        self.memory_usage_factor = (unet_config.get("model_channels", 2048) / 2048) * 0.95
 
     def get_model(self, state_dict, prefix="", device=None):
         out = model_base.CosmosPredict2(self, device=device)
@@ -1026,7 +1026,7 @@ class ZImage(Lumina2):
         "shift": 3.0,
     }
 
-    memory_usage_factor = 1.7
+    memory_usage_factor = 2.0
 
     supported_inference_dtypes = [torch.bfloat16, torch.float16, torch.float32]
 
@@ -1289,7 +1289,7 @@ class ChromaRadiance(Chroma):
     latent_format = comfy.latent_formats.ChromaRadiance
 
     # Pixel-space model, no spatial compression for model input.
-    memory_usage_factor = 0.038
+    memory_usage_factor = 0.044
 
     def get_model(self, state_dict, prefix="", device=None):
         return model_base.ChromaRadiance(self, device=device)
@@ -1332,7 +1332,7 @@ class Omnigen2(supported_models_base.BASE):
         "shift": 2.6,
     }
 
-    memory_usage_factor = 1.65 #TODO
+    memory_usage_factor = 1.95 #TODO
 
     unet_extra_config = {}
     latent_format = latent_formats.Flux
@@ -1397,7 +1397,7 @@ class HunyuanImage21(HunyuanVideo):
 
     latent_format = latent_formats.HunyuanImage21
 
-    memory_usage_factor = 7.7
+    memory_usage_factor = 8.7
 
     supported_inference_dtypes = [torch.bfloat16, torch.float32]
 
@@ -1488,7 +1488,7 @@ class Kandinsky5(supported_models_base.BASE):
     unet_extra_config = {}
     latent_format = latent_formats.HunyuanVideo
 
-    memory_usage_factor = 1.1 #TODO
+    memory_usage_factor = 1.25 #TODO
 
     supported_inference_dtypes = [torch.bfloat16, torch.float32]
 
@@ -1517,7 +1517,7 @@ class Kandinsky5Image(Kandinsky5):
     }
 
     latent_format = latent_formats.Flux
-    memory_usage_factor = 1.1 #TODO
+    memory_usage_factor = 1.25 #TODO
 
     def get_model(self, state_dict, prefix="", device=None):
         out = model_base.Kandinsky5Image(self, device=device)

comfy_api_nodes/nodes_kling.py

@@ -2056,7 +2056,7 @@ class KlingExtension(ComfyExtension):
             OmniProImageToVideoNode,
             OmniProVideoToVideoNode,
             OmniProEditVideoNode,
-            # OmniProImageNode,  # need support from backend
+            OmniProImageNode,
         ]