[Type] Mark input options NotRequired

app/frontend_management.py

@@ -184,27 +184,6 @@ comfyui-frontend-package is not installed.
             )
             sys.exit(-1)
 
-    @classmethod
-    def templates_path(cls) -> str:
-        try:
-            import comfyui_workflow_templates
-
-            return str(
-                importlib.resources.files(comfyui_workflow_templates) / "templates"
-            )
-        except ImportError:
-            logging.error(
-                f"""
-********** ERROR ***********
-
-comfyui-workflow-templates is not installed.
-
-{frontend_install_warning_message()}
-
-********** ERROR ***********
-""".strip()
-            )
-
     @classmethod
     def parse_version_string(cls, value: str) -> tuple[str, str, str]:
         """

comfy/ldm/hidream/model.py

@@ -8,12 +8,25 @@ from einops import repeat
 from comfy.ldm.lightricks.model import TimestepEmbedding, Timesteps
 import torch.nn.functional as F
 
-from comfy.ldm.flux.math import apply_rope, rope
-from comfy.ldm.flux.layers import LastLayer
-
+from comfy.ldm.flux.math import apply_rope
 from comfy.ldm.modules.attention import optimized_attention
 import comfy.model_management
-import comfy.ldm.common_dit
+
+# Copied from https://github.com/black-forest-labs/flux/blob/main/src/flux/math.py
+def rope(pos: torch.Tensor, dim: int, theta: int) -> torch.Tensor:
+    assert dim % 2 == 0, "The dimension must be even."
+
+    scale = torch.arange(0, dim, 2, dtype=torch.float64, device=pos.device) / dim
+    omega = 1.0 / (theta**scale)
+
+    batch_size, seq_length = pos.shape
+    out = torch.einsum("...n,d->...nd", pos, omega)
+    cos_out = torch.cos(out)
+    sin_out = torch.sin(out)
+
+    stacked_out = torch.stack([cos_out, -sin_out, sin_out, cos_out], dim=-1)
+    out = stacked_out.view(batch_size, -1, dim // 2, 2, 2)
+    return out.float()
 
 
 # Copied from https://github.com/black-forest-labs/flux/blob/main/src/flux/modules/layers.py
@@ -71,6 +84,23 @@ class TimestepEmbed(nn.Module):
         return t_emb
 
 
+class OutEmbed(nn.Module):
+    def __init__(self, hidden_size, patch_size, out_channels, dtype=None, device=None, operations=None):
+        super().__init__()
+        self.norm_final = operations.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6, dtype=dtype, device=device)
+        self.linear = operations.Linear(hidden_size, patch_size * patch_size * out_channels, bias=True, dtype=dtype, device=device)
+        self.adaLN_modulation = nn.Sequential(
+            nn.SiLU(),
+            operations.Linear(hidden_size, 2 * hidden_size, bias=True, dtype=dtype, device=device)
+        )
+
+    def forward(self, x, adaln_input):
+        shift, scale = self.adaLN_modulation(adaln_input).chunk(2, dim=1)
+        x = self.norm_final(x) * (1 + scale.unsqueeze(1)) + shift.unsqueeze(1)
+        x = self.linear(x)
+        return x
+
+
 def attention(query: torch.Tensor, key: torch.Tensor, value: torch.Tensor):
     return optimized_attention(query.view(query.shape[0], -1, query.shape[-1] * query.shape[-2]), key.view(key.shape[0], -1, key.shape[-1] * key.shape[-2]), value.view(value.shape[0], -1, value.shape[-1] * value.shape[-2]), query.shape[2])
 
@@ -633,7 +663,7 @@ class HiDreamImageTransformer2DModel(nn.Module):
             ]
         )
 
-        self.final_layer = LastLayer(self.inner_dim, patch_size, self.out_channels, dtype=dtype, device=device, operations=operations)
+        self.final_layer = OutEmbed(self.inner_dim, patch_size, self.out_channels, dtype=dtype, device=device, operations=operations)
 
         caption_channels = [caption_channels[1], ] * (num_layers + num_single_layers) + [caption_channels[0], ]
         caption_projection = []
@@ -702,8 +732,7 @@ class HiDreamImageTransformer2DModel(nn.Module):
         control = None,
         transformer_options = {},
     ) -> torch.Tensor:
-        bs, c, h, w = x.shape
-        hidden_states = comfy.ldm.common_dit.pad_to_patch_size(x, (self.patch_size, self.patch_size))
+        hidden_states = x
         timesteps = t
         pooled_embeds = y
         T5_encoder_hidden_states = context
@@ -796,4 +825,4 @@ class HiDreamImageTransformer2DModel(nn.Module):
         hidden_states = hidden_states[:, :image_tokens_seq_len, ...]
         output = self.final_layer(hidden_states, adaln_input)
         output = self.unpatchify(output, img_sizes)
-        return -output[:, :, :h, :w]
+        return -output

comfy/ldm/wan/model.py

@@ -83,7 +83,7 @@ class WanSelfAttention(nn.Module):
 
 class WanT2VCrossAttention(WanSelfAttention):
 
-    def forward(self, x, context, **kwargs):
+    def forward(self, x, context):
         r"""
         Args:
             x(Tensor): Shape [B, L1, C]
@@ -116,14 +116,14 @@ class WanI2VCrossAttention(WanSelfAttention):
         # self.alpha = nn.Parameter(torch.zeros((1, )))
         self.norm_k_img = RMSNorm(dim, eps=eps, elementwise_affine=True, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")) if qk_norm else nn.Identity()
 
-    def forward(self, x, context, context_img_len):
+    def forward(self, x, context):
         r"""
         Args:
             x(Tensor): Shape [B, L1, C]
             context(Tensor): Shape [B, L2, C]
         """
-        context_img = context[:, :context_img_len]
-        context = context[:, context_img_len:]
+        context_img = context[:, :257]
+        context = context[:, 257:]
 
         # compute query, key, value
         q = self.norm_q(self.q(x))
@@ -193,7 +193,6 @@ class WanAttentionBlock(nn.Module):
         e,
         freqs,
         context,
-        context_img_len=257,
     ):
         r"""
         Args:
@@ -214,7 +213,7 @@ class WanAttentionBlock(nn.Module):
         x = x + y * e[2]
 
         # cross-attention & ffn
-        x = x + self.cross_attn(self.norm3(x), context, context_img_len=context_img_len)
+        x = x + self.cross_attn(self.norm3(x), context)
         y = self.ffn(self.norm2(x) * (1 + e[4]) + e[3])
         x = x + y * e[5]
         return x
@@ -251,7 +250,7 @@ class Head(nn.Module):
 
 class MLPProj(torch.nn.Module):
 
-    def __init__(self, in_dim, out_dim, flf_pos_embed_token_number=None, operation_settings={}):
+    def __init__(self, in_dim, out_dim, operation_settings={}):
         super().__init__()
 
         self.proj = torch.nn.Sequential(
@@ -259,15 +258,7 @@ class MLPProj(torch.nn.Module):
             torch.nn.GELU(), operation_settings.get("operations").Linear(in_dim, out_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")),
             operation_settings.get("operations").LayerNorm(out_dim, device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))
 
-        if flf_pos_embed_token_number is not None:
-            self.emb_pos = nn.Parameter(torch.empty((1, flf_pos_embed_token_number, in_dim), device=operation_settings.get("device"), dtype=operation_settings.get("dtype")))
-        else:
-            self.emb_pos = None
-
     def forward(self, image_embeds):
-        if self.emb_pos is not None:
-            image_embeds = image_embeds[:, :self.emb_pos.shape[1]] + comfy.model_management.cast_to(self.emb_pos[:, :image_embeds.shape[1]], dtype=image_embeds.dtype, device=image_embeds.device)
-
         clip_extra_context_tokens = self.proj(image_embeds)
         return clip_extra_context_tokens
 
@@ -293,7 +284,6 @@ class WanModel(torch.nn.Module):
                  qk_norm=True,
                  cross_attn_norm=True,
                  eps=1e-6,
-                 flf_pos_embed_token_number=None,
                  image_model=None,
                  device=None,
                  dtype=None,
@@ -383,7 +373,7 @@ class WanModel(torch.nn.Module):
         self.rope_embedder = EmbedND(dim=d, theta=10000.0, axes_dim=[d - 4 * (d // 6), 2 * (d // 6), 2 * (d // 6)])
 
         if model_type == 'i2v':
-            self.img_emb = MLPProj(1280, dim, flf_pos_embed_token_number=flf_pos_embed_token_number, operation_settings=operation_settings)
+            self.img_emb = MLPProj(1280, dim, operation_settings=operation_settings)
         else:
             self.img_emb = None
 
@@ -430,12 +420,9 @@ class WanModel(torch.nn.Module):
         # context
         context = self.text_embedding(context)
 
-        context_img_len = None
-        if clip_fea is not None:
-            if self.img_emb is not None:
-                context_clip = self.img_emb(clip_fea)  # bs x 257 x dim
-                context = torch.concat([context_clip, context], dim=1)
-            context_img_len = clip_fea.shape[-2]
+        if clip_fea is not None and self.img_emb is not None:
+            context_clip = self.img_emb(clip_fea)  # bs x 257 x dim
+            context = torch.concat([context_clip, context], dim=1)
 
         patches_replace = transformer_options.get("patches_replace", {})
         blocks_replace = patches_replace.get("dit", {})
@@ -443,12 +430,12 @@ class WanModel(torch.nn.Module):
             if ("double_block", i) in blocks_replace:
                 def block_wrap(args):
                     out = {}
-                    out["img"] = block(args["img"], context=args["txt"], e=args["vec"], freqs=args["pe"], context_img_len=context_img_len)
+                    out["img"] = block(args["img"], context=args["txt"], e=args["vec"], freqs=args["pe"])
                     return out
                 out = blocks_replace[("double_block", i)]({"img": x, "txt": context, "vec": e0, "pe": freqs}, {"original_block": block_wrap})
                 x = out["img"]
             else:
-                x = block(x, e=e0, freqs=freqs, context=context, context_img_len=context_img_len)
+                x = block(x, e=e0, freqs=freqs, context=context)
 
         # head
         x = self.head(x, e)

comfy/model_detection.py

@@ -321,9 +321,6 @@ def detect_unet_config(state_dict, key_prefix, metadata=None):
             dit_config["model_type"] = "i2v"
         else:
             dit_config["model_type"] = "t2v"
-        flf_weight = state_dict.get('{}img_emb.emb_pos'.format(key_prefix))
-        if flf_weight is not None:
-            dit_config["flf_pos_embed_token_number"] = flf_weight.shape[1]
         return dit_config
 
     if '{}latent_in.weight'.format(key_prefix) in state_dict_keys:  # Hunyuan 3D

comfy/rmsnorm.py

@@ -49,7 +49,6 @@ if RMSNorm is None:
                 )
             else:
                 self.register_parameter("weight", None)
-            self.bias = None
 
         def forward(self, x):
             return rms_norm(x, self.weight, self.eps)

comfy/text_encoders/hidream.py

@@ -11,15 +11,14 @@ class HiDreamTokenizer:
     def __init__(self, embedding_directory=None, tokenizer_data={}):
         self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
         self.clip_g = sdxl_clip.SDXLClipGTokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
-        self.t5xxl = sd3_clip.T5XXLTokenizer(embedding_directory=embedding_directory, min_length=128, max_length=128, tokenizer_data=tokenizer_data)
+        self.t5xxl = sd3_clip.T5XXLTokenizer(embedding_directory=embedding_directory, min_length=128, tokenizer_data=tokenizer_data)
         self.llama = hunyuan_video.LLAMA3Tokenizer(embedding_directory=embedding_directory, min_length=128, pad_token=128009, tokenizer_data=tokenizer_data)
 
     def tokenize_with_weights(self, text:str, return_word_ids=False, **kwargs):
         out = {}
         out["g"] = self.clip_g.tokenize_with_weights(text, return_word_ids)
         out["l"] = self.clip_l.tokenize_with_weights(text, return_word_ids)
-        t5xxl = self.t5xxl.tokenize_with_weights(text, return_word_ids)
-        out["t5xxl"] = [t5xxl[0]]  # Use only first 128 tokens
+        out["t5xxl"] = self.t5xxl.tokenize_with_weights(text, return_word_ids)
         out["llama"] = self.llama.tokenize_with_weights(text, return_word_ids)
         return out
 

comfy/text_encoders/sd3_clip.py

@@ -32,9 +32,9 @@ def t5_xxl_detect(state_dict, prefix=""):
     return out
 
 class T5XXLTokenizer(sd1_clip.SDTokenizer):
-    def __init__(self, embedding_directory=None, tokenizer_data={}, min_length=77, max_length=99999999):
+    def __init__(self, embedding_directory=None, tokenizer_data={}, min_length=77):
         tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "t5_tokenizer")
-        super().__init__(tokenizer_path, embedding_directory=embedding_directory, pad_with_end=False, embedding_size=4096, embedding_key='t5xxl', tokenizer_class=T5TokenizerFast, has_start_token=False, pad_to_max_length=False, max_length=max_length, min_length=min_length, tokenizer_data=tokenizer_data)
+        super().__init__(tokenizer_path, embedding_directory=embedding_directory, pad_with_end=False, embedding_size=4096, embedding_key='t5xxl', tokenizer_class=T5TokenizerFast, has_start_token=False, pad_to_max_length=False, max_length=99999999, min_length=min_length, tokenizer_data=tokenizer_data)
 
 
 class SD3Tokenizer:

comfy_extras/nodes_wan.py

@@ -4,7 +4,6 @@ import torch
 import comfy.model_management
 import comfy.utils
 import comfy.latent_formats
-import comfy.clip_vision
 
 
 class WanImageToVideo:
@@ -100,72 +99,6 @@ class WanFunControlToVideo:
         out_latent["samples"] = latent
         return (positive, negative, out_latent)
 
-class WanFirstLastFrameToVideo:
-    @classmethod
-    def INPUT_TYPES(s):
-        return {"required": {"positive": ("CONDITIONING", ),
-                             "negative": ("CONDITIONING", ),
-                             "vae": ("VAE", ),
-                             "width": ("INT", {"default": 832, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
-                             "height": ("INT", {"default": 480, "min": 16, "max": nodes.MAX_RESOLUTION, "step": 16}),
-                             "length": ("INT", {"default": 81, "min": 1, "max": nodes.MAX_RESOLUTION, "step": 4}),
-                             "batch_size": ("INT", {"default": 1, "min": 1, "max": 4096}),
-                },
-                "optional": {"clip_vision_start_image": ("CLIP_VISION_OUTPUT", ),
-                             "clip_vision_end_image": ("CLIP_VISION_OUTPUT", ),
-                             "start_image": ("IMAGE", ),
-                             "end_image": ("IMAGE", ),
-                }}
-
-    RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT")
-    RETURN_NAMES = ("positive", "negative", "latent")
-    FUNCTION = "encode"
-
-    CATEGORY = "conditioning/video_models"
-
-    def encode(self, positive, negative, vae, width, height, length, batch_size, start_image=None, end_image=None, clip_vision_start_image=None, clip_vision_end_image=None):
-        latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
-        if start_image is not None:
-            start_image = comfy.utils.common_upscale(start_image[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
-        if end_image is not None:
-            end_image = comfy.utils.common_upscale(end_image[-length:].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
-
-        image = torch.ones((length, height, width, 3)) * 0.5
-        mask = torch.ones((1, 1, latent.shape[2] * 4, latent.shape[-2], latent.shape[-1]))
-
-        if start_image is not None:
-            image[:start_image.shape[0]] = start_image
-            mask[:, :, :start_image.shape[0] + 3] = 0.0
-
-        if end_image is not None:
-            image[-end_image.shape[0]:] = end_image
-            mask[:, :, -end_image.shape[0]:] = 0.0
-
-        concat_latent_image = vae.encode(image[:, :, :, :3])
-        mask = mask.view(1, mask.shape[2] // 4, 4, mask.shape[3], mask.shape[4]).transpose(1, 2)
-        positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
-        negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
-
-        if clip_vision_start_image is not None:
-            clip_vision_output = clip_vision_start_image
-
-        if clip_vision_end_image is not None:
-            if clip_vision_output is not None:
-                states = torch.cat([clip_vision_output.penultimate_hidden_states, clip_vision_end_image.penultimate_hidden_states], dim=-2)
-                clip_vision_output = comfy.clip_vision.Output()
-                clip_vision_output.penultimate_hidden_states = states
-            else:
-                clip_vision_output = clip_vision_end_image
-
-        if clip_vision_output is not None:
-            positive = node_helpers.conditioning_set_values(positive, {"clip_vision_output": clip_vision_output})
-            negative = node_helpers.conditioning_set_values(negative, {"clip_vision_output": clip_vision_output})
-
-        out_latent = {}
-        out_latent["samples"] = latent
-        return (positive, negative, out_latent)
-
-
 class WanFunInpaintToVideo:
     @classmethod
     def INPUT_TYPES(s):
@@ -189,13 +122,38 @@ class WanFunInpaintToVideo:
     CATEGORY = "conditioning/video_models"
 
     def encode(self, positive, negative, vae, width, height, length, batch_size, start_image=None, end_image=None, clip_vision_output=None):
-        flfv = WanFirstLastFrameToVideo()
-        return flfv.encode(positive, negative, vae, width, height, length, batch_size, start_image=start_image, end_image=end_image, clip_vision_start_image=clip_vision_output)
+        latent = torch.zeros([batch_size, 16, ((length - 1) // 4) + 1, height // 8, width // 8], device=comfy.model_management.intermediate_device())
+        if start_image is not None:
+            start_image = comfy.utils.common_upscale(start_image[:length].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
+        if end_image is not None:
+            end_image = comfy.utils.common_upscale(end_image[-length:].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)
 
+        image = torch.ones((length, height, width, 3)) * 0.5
+        mask = torch.ones((1, 1, latent.shape[2] * 4, latent.shape[-2], latent.shape[-1]))
+
+        if start_image is not None:
+            image[:start_image.shape[0]] = start_image
+            mask[:, :, :start_image.shape[0] + 3] = 0.0
+
+        if end_image is not None:
+            image[-end_image.shape[0]:] = end_image
+            mask[:, :, -end_image.shape[0]:] = 0.0
+
+        concat_latent_image = vae.encode(image[:, :, :, :3])
+        mask = mask.view(1, mask.shape[2] // 4, 4, mask.shape[3], mask.shape[4]).transpose(1, 2)
+        positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
+        negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": concat_latent_image, "concat_mask": mask})
+
+        if clip_vision_output is not None:
+            positive = node_helpers.conditioning_set_values(positive, {"clip_vision_output": clip_vision_output})
+            negative = node_helpers.conditioning_set_values(negative, {"clip_vision_output": clip_vision_output})
+
+        out_latent = {}
+        out_latent["samples"] = latent
+        return (positive, negative, out_latent)
 
 NODE_CLASS_MAPPINGS = {
     "WanImageToVideo": WanImageToVideo,
     "WanFunControlToVideo": WanFunControlToVideo,
     "WanFunInpaintToVideo": WanFunInpaintToVideo,
-    "WanFirstLastFrameToVideo": WanFirstLastFrameToVideo,
 }

comfyui_version.py

@@ -1,3 +1,3 @@
 # This file is automatically generated by the build process when version is
 # updated in pyproject.toml.
-__version__ = "0.3.29"
+__version__ = "0.3.28"

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ComfyUI"
-version = "0.3.29"
+version = "0.3.28"
 readme = "README.md"
 license = { file = "LICENSE" }
 requires-python = ">=3.9"

requirements.txt

@@ -1,5 +1,4 @@
-comfyui-frontend-package==1.16.8
-comfyui-workflow-templates==0.1.1
+comfyui-frontend-package==1.15.13
 torch
 torchsde
 torchvision

server.py

@@ -736,12 +736,6 @@ class PromptServer():
         for name, dir in nodes.EXTENSION_WEB_DIRS.items():
             self.app.add_routes([web.static('/extensions/' + name, dir)])
 
-        workflow_templates_path = FrontendManager.templates_path()
-        if workflow_templates_path:
-            self.app.add_routes([
-                web.static('/templates', workflow_templates_path)
-            ])
-
         self.app.add_routes([
             web.static('/', self.web_root),
         ])