feat: add performance hyper-sd based on 4step LoRA (#2812)

* feat: add performance hyper-sd based on 4step LoRA

* feat: use LoRA weight 0.8, sampler dpmpp_sde_gpu and scheduler_name karras

suggested in https://github.com/lllyasviel/Fooocus/discussions/2813#discussioncomment-9245251
results see https://github.com/lllyasviel/Fooocus/discussions/2813#discussioncomment-9275251

* feat: change ByteDance huggingface profile with mashb1t

* wip: add hyper-sd 8 step cfg lora with negative prompt support

* feat: remove hyper-sd8 performance

still waiting for the release of hyper-sd 4step CFG LoRA, not yet satisfied with any of the CFG LoRAs compared to non-cfg ones.
see https://huggingface.co/ByteDance/Hyper-SD

modules/async_worker.py

@@ -275,6 +275,33 @@ def worker():
             adm_scaler_negative = 1.0
             adm_scaler_end = 0.0
 
+        elif performance_selection == Performance.HYPER_SD:
+            print('Enter Hyper-SD mode.')
+            progressbar(async_task, 1, 'Downloading Hyper-SD components ...')
+            loras += [(modules.config.downloading_sdxl_hyper_sd_lora(), 0.8)]
+
+            if refiner_model_name != 'None':
+                print(f'Refiner disabled in Hyper-SD mode.')
+
+            refiner_model_name = 'None'
+            sampler_name = 'dpmpp_sde_gpu'
+            scheduler_name = 'karras'
+            sharpness = 0.0
+            guidance_scale = 1.0
+            adaptive_cfg = 1.0
+            refiner_switch = 1.0
+            adm_scaler_positive = 1.0
+            adm_scaler_negative = 1.0
+            adm_scaler_end = 0.0
+
+        elif performance_selection == Performance.HYPER_SD8:
+            print('Enter Hyper-SD8 mode.')
+            progressbar(async_task, 1, 'Downloading Hyper-SD components ...')
+            loras += [(modules.config.downloading_sdxl_hyper_sd_cfg_lora(), 0.3)]
+
+            sampler_name = 'dpmpp_sde_gpu'
+            scheduler_name = 'normal'
+
         print(f'[Parameters] Adaptive CFG = {adaptive_cfg}')
         print(f'[Parameters] Sharpness = {sharpness}')
         print(f'[Parameters] ControlNet Softness = {controlnet_softness}')

modules/config.py

@@ -553,7 +553,8 @@ wildcard_filenames = []
 
 sdxl_lcm_lora = 'sdxl_lcm_lora.safetensors'
 sdxl_lightning_lora = 'sdxl_lightning_4step_lora.safetensors'
-loras_metadata_remove = [sdxl_lcm_lora, sdxl_lightning_lora]
+sdxl_hyper_sd_lora = 'sdxl_hyper_sd_4step_lora.safetensors'
+loras_metadata_remove = [sdxl_lcm_lora, sdxl_lightning_lora, sdxl_hyper_sd_lora]
 
 
 def get_model_filenames(folder_paths, extensions=None, name_filter=None):
@@ -627,13 +628,22 @@ def downloading_sdxl_lcm_lora():
 
 def downloading_sdxl_lightning_lora():
     load_file_from_url(
-        url='https://huggingface.co/ByteDance/SDXL-Lightning/resolve/main/sdxl_lightning_4step_lora.safetensors',
+        url='https://huggingface.co/mashb1t/misc/resolve/main/sdxl_lightning_4step_lora.safetensors',
         model_dir=paths_loras[0],
         file_name=sdxl_lightning_lora
     )
     return sdxl_lightning_lora
 
 
+def downloading_sdxl_hyper_sd_lora():
+    load_file_from_url(
+        url='https://huggingface.co/mashb1t/misc/resolve/main/sdxl_hyper_sd_4step_lora.safetensors',
+        model_dir=paths_loras[0],
+        file_name=sdxl_hyper_sd_lora
+    )
+    return sdxl_hyper_sd_lora
+
+
 def downloading_controlnet_canny():
     load_file_from_url(
         url='https://huggingface.co/lllyasviel/misc/resolve/main/control-lora-canny-rank128.safetensors',

modules/flags.py

@@ -110,6 +110,7 @@ class Steps(IntEnum):
     SPEED = 30
     EXTREME_SPEED = 8
     LIGHTNING = 4
+    HYPER_SD = 4
 
 
 class StepsUOV(IntEnum):
@@ -117,6 +118,7 @@ class StepsUOV(IntEnum):
     SPEED = 18
     EXTREME_SPEED = 8
     LIGHTNING = 4
+    HYPER_SD = 4
 
 
 class Performance(Enum):
@@ -124,6 +126,7 @@ class Performance(Enum):
     SPEED = 'Speed'
     EXTREME_SPEED = 'Extreme Speed'
     LIGHTNING = 'Lightning'
+    HYPER_SD = 'Hyper-SD'
 
     @classmethod
     def list(cls) -> list:
@@ -133,7 +136,7 @@ class Performance(Enum):
     def has_restricted_features(cls, x) -> bool:
         if isinstance(x, Performance):
             x = x.value
-        return x in [cls.EXTREME_SPEED.value, cls.LIGHTNING.value]
+        return x in [cls.EXTREME_SPEED.value, cls.LIGHTNING.value, cls.HYPER_SD.value]
 
     def steps(self) -> int | None:
         return Steps[self.name].value if Steps[self.name] else None