enable sfast in controlnet pipeline

build-run.sh

@@ -13,4 +13,4 @@ if [ -z ${PIPELINE+x} ]; then
     PIPELINE="controlnet"
 fi
 echo -e "\033[1;32m\npipeline: $PIPELINE \033[0m"
-python3 run.py --port 7860 --host 0.0.0.0 --pipeline $PIPELINE
+python3 run.py --port 7860 --host 0.0.0.0 --pipeline $PIPELINE --sfast

pipelines/controlnet.py

@@ -173,16 +173,24 @@ class Pipeline:
             self.pipe.vae = AutoencoderTiny.from_pretrained(
                 taesd_model, torch_dtype=torch_dtype, use_safetensors=True
             ).to(device)
+
+        if args.sfast:
+            from sfast.compilers.stable_diffusion_pipeline_compiler import (
+                compile,
+                CompilationConfig,
+            )
+
+            config = CompilationConfig.Default()
+            config.enable_xformers = True
+            config.enable_triton = True
+            config.enable_cuda_graph = True
+            self.pipe = compile(self.pipe, config=config)
         self.canny_torch = SobelOperator(device=device)
         self.pipe.set_progress_bar_config(disable=True)
         self.pipe.to(device=device, dtype=torch_dtype)
         if device.type != "mps":
             self.pipe.unet.to(memory_format=torch.channels_last)
 
-        # check if computer has less than 64GB of RAM using sys or os
-        if psutil.virtual_memory().total < 64 * 1024**3:
-            self.pipe.enable_attention_slicing()
-
         if args.torch_compile:
             self.pipe.unet = torch.compile(
                 self.pipe.unet, mode="reduce-overhead", fullgraph=True