diffusers#4003 Subtract one from the begin step to match the ending step, due to the off-by-one count issue

Author: bghira

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py

@@ -887,6 +887,7 @@ def __call__(
             # In Diffusers pipelines, each generation always does ONE more step than asked.
             # As a result, we need to reduce the number of steps we do for this subsequent by one, to remain aligned.
             num_inference_steps -= 1
+            begin_inference_step -= 1
         self.scheduler.set_timesteps(num_inference_steps, device=device)
         timesteps, num_inference_steps = self.get_timesteps(num_inference_steps, begin_inference_step, strength, device)
         latent_timestep = timesteps[:1].repeat(batch_size * num_images_per_prompt)