add good

app.py

@@ -14,6 +14,7 @@ from diffusers import (
     T2IAdapter,
 )
 import time
+import utils
 
 
 dtype = torch.float16
@@ -138,12 +139,28 @@ def generate(
     for _ in range(3):
         prompt = 77 * "a"
         num_inference_steps = 20
-        start_time = time.time()
-        _ = pipeline(
-            prompt,
+        call_args = dict(
+            prompt=prompt,
             num_images_per_prompt=num_images_per_prompt,
             num_inference_steps=num_inference_steps,
-        ).images
+        )
+
+        if pipeline_to_benchmark in ["SD I2I", "SDXL I2I"]:
+            image = utils.get_image_for_img_to_img(pipeline_to_benchmark)
+            call_args.update({"image": image})
+        elif "Inpainting" in pipeline_to_benchmark:
+            image, mask_image = utils.get_image_from_inpainting(pipeline_to_benchmark)
+            call_args.update({"image": image, "mask_image": mask_image})
+        elif "ControlNet" in pipeline_to_benchmark:
+            image = utils.get_image_for_controlnet(pipeline_to_benchmark)
+            call_args.update({"image": image})
+
+        elif "Adapters" in pipeline_to_benchmark:
+            image = utils.get_image_for_adapters(pipeline_to_benchmark)
+            call_args.update({"image": image})
+
+        start_time = time.time()
+        _ = pipeline(**call_args).images
         end_time = time.time()
 
         print(f"For {num_inference_steps} steps", end_time - start_time)

utils.py

@@ -0,0 +1,52 @@
+from diffusers.utils import load_image
+
+
+def get_image_for_img_to_img(pipeline_to_benchmark):
+    if pipeline_to_benchmark == "SD I2I":
+        url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
+        init_image = load_image(url).convert("RGB")
+        size = (768, 512)
+        init_image = init_image.resize(size)
+    elif pipeline_to_benchmark == "SDXL I2I":
+        url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-img2img.png"
+        init_image = load_image(url).convert("RGB")
+    return init_image
+
+
+def get_image_from_inpainting(pipeline_to_benchmark):
+    if pipeline_to_benchmark == "SD I2I":
+        image_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo.png"
+        mask_url = "https://raw.githubusercontent.com/CompVis/latent-diffusion/main/data/inpainting_examples/overture-creations-5sI6fQgYIuo_mask.png"
+        init_image = load_image(image_url).convert("RGB").resize((512, 512))
+        mask_image = load_image(mask_url).convert("RGB").resize((512, 512))
+    elif pipeline_to_benchmark == "SDXL I2I":
+        image_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-text2img.png"
+        mask_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/sdxl-inpaint-mask.png"
+        init_image = load_image(image_url).convert("RGB")
+        mask_image = load_image(mask_url).convert("RGB")
+    return init_image, mask_image
+
+
+def get_image_for_controlnet(pipeline_to_benchmark):
+    if pipeline_to_benchmark == "SD ControlNet":
+        image = load_image(
+            "https://huggingface.co/lllyasviel/sd-controlnet-hed/resolve/main/images/bird.png"
+        )
+    elif pipeline_to_benchmark == "SDXL ControlNet":
+        image = load_image(
+            "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/sd_controlnet/hf-logo.png"
+        )
+
+    return image
+
+
+def get_image_for_adapters(pipeline_to_benchmark):
+    if pipeline_to_benchmark == "SD T2I Adapters":
+        image = load_image(
+            "https://huggingface.co/TencentARC/t2iadapter_canny_sd14v1/resolve/main/images/canny.png"
+        ).convert("L")
+    elif pipeline_to_benchmark == "SDXL T2I Adapters":
+        image = load_image(
+            "https://huggingface.co/Adapter/t2iadapter/resolve/main/figs_SDXLV1.0/cond_canny.png"
+        ).convert("L")
+    return image