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

ui/src/types.ts

@@ -71,6 +71,7 @@ export interface NetworkConfig {
   network_kwargs: {
     ignore_if_contains: string[];
   };
+  transformer_only?: boolean;
 }
 
 export interface SaveConfig {
@@ -149,6 +150,7 @@ export interface TrainConfig {
   do_differential_guidance?: boolean;
   differential_guidance_scale?: number;
   audio_loss_multiplier?: number;
+  max_loss?: number | null;
 }
 
 export interface QuantizeKwargsConfig {