libtcmalloc

Dockerfile

@@ -11,7 +11,9 @@ RUN apt-get update && apt-get install --no-install-recommends -y \
     python3-dev \
     git \
     ffmpeg \
+    google-perftools \
     && apt-get clean && rm -rf /var/lib/apt/lists/*
+    
 
 WORKDIR /code
 
@@ -36,5 +38,6 @@ WORKDIR $HOME/app
 # Copy the current directory contents into the container at $HOME/app setting the owner to the user
 COPY --chown=user . $HOME/app
 
+ENV LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc.so.4
 CMD ["uvicorn", "app-img2img:app", "--host", "0.0.0.0", "--port", "7860"]
 # CMD ["uvicorn", "app-txt2img:app", "--host", "0.0.0.0", "--port", "7860"]

app-img2img.py

@@ -102,7 +102,7 @@ def predict(
 ):
     generator = torch.manual_seed(params.seed)
     # Can be set to 1~50 steps. LCM support fast inference even <= 4 steps. Recommend: 1~8 steps.
-    num_inference_steps = 3
+    num_inference_steps = 4
     results = pipe(
         prompt_embeds=prompt_embeds,
         generator=generator,

app-txt2img.py

@@ -100,7 +100,7 @@ def predict(params: InputParams):
     generator = torch.manual_seed(params.seed)
     prompt_embeds = compel_proc(params.prompt)
     # Can be set to 1~50 steps. LCM support fast inference even <= 4 steps. Recommend: 1~8 steps.
-    num_inference_steps = 8
+    num_inference_steps = 4
     results = pipe(
         prompt_embeds=prompt_embeds,
         generator=generator,