Kandinsky5 model support (#10988)

* Add Kandinsky5 model support

lite and pro T2V tested to work

* Update kandinsky5.py

* Fix fp8

* Fix fp8_scaled text encoder

* Add transformer_options for attention

* Code cleanup, optimizations, use fp32 for all layers originally at fp32

* ImageToVideo -node

* Fix I2V, add necessary latent post process nodes

* Support text to image model

* Support block replace patches (SLG mostly)

* Support official LoRAs

* Don't scale RoPE for lite model as that just doesn't work...

* Update supported_models.py

* Rever RoPE scaling to simpler one

* Fix typo

* Handle latent dim difference for image model in the VAE instead

* Add node to use different prompts for clip_l and qwen25_7b

* Reduce peak VRAM usage a bit

* Further reduce peak VRAM consumption by chunking ffn

* Update chunking

* Update memory_usage_factor

* Code cleanup, don't force the fp32 layers as it has minimal effect

* Allow for stronger changes with first frames normalization

Default values are too weak for any meaningful changes, these should probably be exposed as advanced node options when that's available.

* Add image model's own chat template, remove unused image2video template

* Remove hard error in ReplaceVideoLatentFrames -node

* Update kandinsky5.py

* Update supported_models.py

* Fix typos in prompt template

They were now fixed in the original repository as well

* Update ReplaceVideoLatentFrames

Add tooltips
Make source optional
Better handle negative index

* Rename NormalizeVideoLatentFrames -node

For bit better clarity what it does

* Fix NormalizeVideoLatentStart node out on non-op

comfy_extras/nodes_latent.py

@@ -4,7 +4,7 @@ import torch
 import nodes
 from typing_extensions import override
 from comfy_api.latest import ComfyExtension, io
-
+import logging
 
 def reshape_latent_to(target_shape, latent, repeat_batch=True):
     if latent.shape[1:] != target_shape[1:]:
@@ -388,6 +388,42 @@ class LatentOperationSharpen(io.ComfyNode):
             return luminance * sharpened
         return io.NodeOutput(sharpen)
 
+class ReplaceVideoLatentFrames(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ReplaceVideoLatentFrames",
+            category="latent/batch",
+            inputs=[
+                io.Latent.Input("destination", tooltip="The destination latent where frames will be replaced."),
+                io.Latent.Input("source", optional=True, tooltip="The source latent providing frames to insert into the destination latent. If not provided, the destination latent is returned unchanged."),
+                io.Int.Input("index", default=0, min=-nodes.MAX_RESOLUTION, max=nodes.MAX_RESOLUTION, step=1, tooltip="The starting latent frame index in the destination latent where the source latent frames will be placed. Negative values count from the end."),
+            ],
+            outputs=[
+                io.Latent.Output(),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, destination, index, source=None) -> io.NodeOutput:
+        if source is None:
+            return io.NodeOutput(destination)
+        dest_frames = destination["samples"].shape[2]
+        source_frames = source["samples"].shape[2]
+        if index < 0:
+            index = dest_frames + index
+        if index > dest_frames:
+            logging.warning(f"ReplaceVideoLatentFrames: Index {index} is out of bounds for destination latent frames {dest_frames}.")
+            return io.NodeOutput(destination)
+        if index + source_frames > dest_frames:
+            logging.warning(f"ReplaceVideoLatentFrames: Source latent frames {source_frames} do not fit within destination latent frames {dest_frames} at the specified index {index}.")
+            return io.NodeOutput(destination)
+        s = source.copy()
+        s_source = source["samples"]
+        s_destination = destination["samples"].clone()
+        s_destination[:, :, index:index + s_source.shape[2]] = s_source
+        s["samples"] = s_destination
+        return io.NodeOutput(s)
 
 class LatentExtension(ComfyExtension):
     @override
@@ -405,6 +441,7 @@ class LatentExtension(ComfyExtension):
             LatentApplyOperationCFG,
             LatentOperationTonemapReinhard,
             LatentOperationSharpen,
+            ReplaceVideoLatentFrames
         ]