Make dpmpp_sde(_k) use not random seed

invokeai/app/invocations/latent.py

@@ -13,7 +13,7 @@ from diffusers.models.attention_processor import (
     LoRAXFormersAttnProcessor,
     XFormersAttnProcessor,
 )
-from diffusers.schedulers import SchedulerMixin as Scheduler
+from diffusers.schedulers import DPMSolverSDEScheduler, SchedulerMixin as Scheduler
 from pydantic import BaseModel, Field, validator
 from torchvision.transforms.functional import resize as tv_resize
 
@@ -81,6 +81,7 @@ def get_scheduler(
     context: InvocationContext,
     scheduler_info: ModelInfo,
     scheduler_name: str,
+    seed: int,
 ) -> Scheduler:
     scheduler_class, scheduler_extra_config = SCHEDULER_MAP.get(scheduler_name, SCHEDULER_MAP["ddim"])
     orig_scheduler_info = context.services.model_manager.get_model(
@@ -97,6 +98,11 @@ def get_scheduler(
         **scheduler_extra_config,
         "_backup": scheduler_config,
     }
+
+    # make dpmpp_sde reproducable(seed can be passed only in initializer)
+    if scheduler_class is DPMSolverSDEScheduler:
+        scheduler_config["noise_sampler_seed"] = seed
+
     scheduler = scheduler_class.from_config(scheduler_config)
 
     # hack copied over from generate.py
@@ -421,6 +427,7 @@ class DenoiseLatentsInvocation(BaseInvocation):
                     context=context,
                     scheduler_info=self.unet.scheduler,
                     scheduler_name=self.scheduler,
+                    seed=seed,
                 )
 
                 pipeline = self.create_pipeline(unet, scheduler)

invokeai/app/invocations/onnx.py

@@ -212,6 +212,7 @@ class ONNXTextToLatentsInvocation(BaseInvocation):
             context=context,
             scheduler_info=self.unet.scheduler,
             scheduler_name=self.scheduler,
+            seed=0, # TODO: refactor this node
         )
 
         def torch2numpy(latent: torch.Tensor):