$ sudo apt-get -y install cuda Reading package lists... Done Building dependency tree Reading state information... Done W: GPG error: file:/var/cuda-repo-wsl-ubuntu-11-7-local InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY BCD50790B81839D3 E: The repository 'file:/var/cuda-repo-wsl-ubuntu-11-7-local InRelease' is not signed. N: Updating from such a repository can't be done securely, and is therefore disabled by default. N: See apt-secure(8) manpage for repository creation and user configuration details.

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600 wget https://developer.download.nvidia.com/compute/cuda/11.7.1/local_installers/cuda-repo-wsl-ubuntu-11-7-local_11.7.1-1_amd64.deb sudo dpkg -i cuda-repo-wsl-ubuntu-11-7-local_11.7.1-1_amd64.deb sudo cp /var/cuda-repo-wsl-ubuntu-11-7-local/cuda-*-keyring.gpg /usr/share/keyrings/ # ↑公式のリファレンスは鍵のインストールもあって親切 sudo apt-get update sudo apt-get -y install cuda

git clone https://github.com/NVIDIA/cuda-samples.git cd uda-samples/Samples/0_Introduction/asyncAPI make run ./asyncAPI [./asyncAPI] - Starting... GPU Device 0: "Pascal" with compute capability 6.1 CUDA device [NVIDIA GeForce GTX 1070] time spent executing by the GPU: 11.05 time spent by CPU in CUDA calls: 1.49

sudo apt-get install python3-venv python3 -m venv env source env/bin/activate # deactivate でもとに戻れる pip install diffusers==0.2.4 transformers scipy ftfy

sudo apt-get install wget ca-certificates

export STABLE_DIFFUSION_TOKEN="/your/huggingface/hub/token"

exec zsh -l

import torch from diffusers import StableDiffusionPipeline import os from torch import autocast # get your token at https://huggingface.co/settings/tokens pipe = StableDiffusionPipeline.from_pretrained( "CompVis/stable-diffusion-v1-4", revision="fp16", torch_dtype=torch.float16, use_auth_token=os.environ["STABLE_DIFFUSION_TOKEN"], ) pipe.to("cuda") prompt = "a photograph of an astronaut riding a horse" # この行は公式のマニュアルと少し違うので注意（公式が不完全） with autocast("cuda"): image = pipe(prompt)["sample"][0] image.save(f"astronaut_rides_horse.png")

python3 main.py

$ python3 main.py /home/motoso/sd/env/lib/python3.8/site-packages/torch/cuda/__init__.py:146: UserWarning: NVIDIA GeForce RTX 3090 with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70. If you want to use the NVIDIA GeForce RTX 3090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

/usr/local/cuda/bin/nvcc --version nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2022 NVIDIA Corporation Built on Wed_Jun__8_16:49:14_PDT_2022 Cuda compilation tools, release 11.7, V11.7.99 Build cuda_11.7.r11.7/compiler.31442593_0

pip uninstall torch pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

python3 main.py 51it [00:05, 8.84it/s]

apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu2004/x86_64/7fa2af80.pub

sudo apt-key del 7fa2af80 sudo apt-key del 3bf863cc