Apply suggestions from code review

Co-authored-by: hlky <[email protected]>

Author: faaany

tests/pipelines/stable_diffusion/test_stable_diffusion_inpaint.py

@@ -709,8 +709,8 @@ def test_stable_diffusion_inpaint_k_lms(self):
 
     def test_stable_diffusion_inpaint_with_sequential_cpu_offloading(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         pipe = StableDiffusionInpaintPipeline.from_pretrained(
             "botp/stable-diffusion-v1-5-inpainting", safety_checker=None, torch_dtype=torch.float16
@@ -912,8 +912,8 @@ def test_stable_diffusion_inpaint_k_lms(self):
 
     def test_stable_diffusion_inpaint_with_sequential_cpu_offloading(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         vae = AsymmetricAutoencoderKL.from_pretrained(
             "cross-attention/asymmetric-autoencoder-kl-x-1-5", torch_dtype=torch.float16

tests/pipelines/stable_diffusion/test_stable_diffusion_instruction_pix2pix.py

@@ -389,8 +389,8 @@ def callback_fn(step: int, timestep: int, latents: torch.Tensor) -> None:
 
     def test_stable_diffusion_pipeline_with_sequential_cpu_offloading(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
             "timbrooks/instruct-pix2pix", safety_checker=None, torch_dtype=torch.float16

tests/pipelines/stable_diffusion_2/test_stable_diffusion_inpaint.py

@@ -252,8 +252,8 @@ def test_stable_diffusion_inpaint_pipeline_fp16(self):
 
     def test_stable_diffusion_pipeline_with_sequential_cpu_offloading(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         init_image = load_image(
             "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main"

tests/pipelines/stable_diffusion_2/test_stable_diffusion_v_pred.py

@@ -535,8 +535,8 @@ def test_stable_diffusion_low_cpu_mem_usage_v_pred(self):
 
     def test_stable_diffusion_pipeline_with_sequential_cpu_offloading_v_pred(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         pipeline_id = "stabilityai/stable-diffusion-2"
         prompt = "Andromeda galaxy in a bottle"

tests/pipelines/stable_diffusion_image_variation/test_stable_diffusion_image_variation.py

@@ -263,8 +263,8 @@ def callback_fn(step: int, timestep: int, latents: torch.Tensor) -> None:
 
     def test_stable_diffusion_pipeline_with_sequential_cpu_offloading(self):
         backend_empty_cache(torch_device)
-        backend_reset_peak_memory_stats(torch_device)
         backend_reset_max_memory_allocated(torch_device)
+        backend_reset_peak_memory_stats(torch_device)
 
         pipe = StableDiffusionImageVariationPipeline.from_pretrained(
             "lambdalabs/sd-image-variations-diffusers", safety_checker=None, torch_dtype=torch.float16