Diffusers Quick Start
Make sure that you have the correct version of python in use and that PyTorch is installed, the correct version for your GPU.
python -m venv venv
. venv/bin/activate
pip3 install torch torchvision --index-url https://download.pytorch.org/whl/rocm6.4
pip3 install diffusers["torch"] accelerate transformers
To run a pipeline, create a python file containing the following, cuda here
is fine for an AMD GPU using ROCm:
from diffusers import DiffusionPipeline
import torch
base_pipeline = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True
)
base_pipeline.to("cuda", torch.float16)
Then build and execute the pipeline:
seed = 43827935
steps = 25
width = 832
height = 1216
positive_prompt = "Colorful Roaring Lion sitting on a victorian couch, flat icon, vector"
negative_prompt = "realistic, realism, photograph"
torch.manual_seed(seed)
generator = torch.Generator(device="cuda").manual_seed(seed)
images = base_pipeline(width=width,
height=height,
guidance_scale=8,
num_inference_steps=steps,
prompt=positive_prompt,
negative_prompt=negative_prompt,
generator=generator).images