Create pipeline.py

pipeline.py

@@ -0,0 +1,21 @@
+import torch
+
+from diffusers import DiffusionPipeline
+
+
+class UnetSchedulerOneForwardPipeline(DiffusionPipeline):
+    def __init__(self, unet, scheduler):
+        super().__init__()
+
+        self.register_modules(unet=unet, scheduler=scheduler)
+
+    def __call__(self):
+        image = torch.randn(
+            (1, self.unet.in_channels, self.unet.sample_size, self.unet.sample_size),
+        )
+        timestep = 1
+
+        model_output = self.unet(image, timestep).sample
+        scheduler_output = self.scheduler.step(model_output, timestep, image).prev_sample
+
+        return scheduler_output