It's very possible that I've done something terribly wrong here. I've tried to make sure it's super easy to reproduce - I've just re-tested the reproduction steps below from scratch to be 100% sure, and am still seeing the same result.
docker run --rm --gpus all -p 5000:5000 r8.im/camenduru/animatediff-lightning-4-step@sha256:be39c6d599942831314b770f03cfd062bfd0faa8cc52e9289bcce830b721fcb6
Then run this:
curl -s -X POST \
-H "Content-Type: application/json" \
-d $'{
"input": {
"prompt": "A dog running around",
"guidance_scale": 1
}
}' \
http://localhost:5000/predictions
Test A1111 speed (~1 it/s on 4090):
docker run -p 7860:7860 -it --rm --gpus all pytorch/pytorch:2.2.1-cuda12.1-cudnn8-devel bash
apt update
# install a1111 (according to readme):
apt install wget git python3 python3-venv libgl1 libglib2.0-0 -y
wget -q https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/webui.sh
chmod +x ./webui.sh
./webui.sh -f --listen --xformers # wait for install, then exit
# install sd-webui-animatediff:
cd stable-diffusion-webui/extensions
git clone https://github.com/continue-revolution/sd-webui-animatediff
cd sd-webui-animatediff/model
wget -O "animatediff_lightning_4step_comfyui.safetensors" https://huggingface.co/ByteDance/AnimateDiff-Lightning/resolve/main/animatediff_lightning_4step_comfyui.safetensors?download=true
cd /workspace
./webui.sh -f --listen --xformers
# Leave everything default except:
# Prompt = A dog running around
# CFG Scale = 1
# Sampling steps = 4
# Sampling method = DPM++SDE
# Enable AnimateDiff = checked
# Number of frames = 16
It's very possible that I've done something terribly wrong here. I've tried to make sure it's super easy to reproduce - I've just re-tested the reproduction steps below from scratch to be 100% sure, and am still seeing the same result.
AUTOMATIC1111/stable-diffusion-webui
:version: v1.8.0 • python: 3.10.13 • torch: 2.1.2+cu121 • xformers: 0.0.23.post1 • gradio: 3.41.2
continue-revolution/sd-webui-animatediff
:Reproduction steps:
Test official ByteDance Gradio demo speed (~5 it/s on 4090):
Run the docker command below, then open
localhost:7860
and type a prompt:Test ComfyUI demo speed (~5 it/s on 4090):
Run this docker command:
Then run this:
Test A1111 speed (~1 it/s on 4090):