Add support for fine-tuning Hidream O1 (#831)

* Initial support for hidream. Lora keys likely need work

* Fix saving for hidream-o1

* Remove dependence on flash attention for hidream o1

* Fix gradient checkpointing for hidream o1

* A lot of fixes for hidream. Handle loading and saving as comfy model.

* Omit layers not used in comfy. Fix issue with lora loading keys in comfy

* Version bumpo

extensions_built_in/diffusion_models/__init__.py

@@ -11,6 +11,7 @@ from .ltx2 import LTX2Model, LTX23Model
 from .zeta_chroma import ZetaChromaModel
 from .ernie_image import ErnieImageModel
 from .nucleus_image import NucleusImageModel
+from .hidream.hidream_o1_model import HidreamO1Model
 
 AI_TOOLKIT_MODELS = [
     # put a list of models here
@@ -36,4 +37,5 @@ AI_TOOLKIT_MODELS = [
     ZetaChromaModel,
     ErnieImageModel,
     NucleusImageModel,
+    HidreamO1Model,
 ]