AB型技術系 主に備忘録

ほぼプログラム関連の備忘録

NVIDIA GeForce GT 710でStable Diffusion WebUIは当然動かない

生成AIを試したくてStable Diffusion WebUIをインストールした時のメモ書きです。

Stable Diffusion WebUI

stablediffusionweb.com

PCスペック

  • OS:Windows10
  • CPU:Ryzen 9 3900X
  • GPU:玄人志向 NVIDIA GeForce GT 710
  • VRAM:1GB
  • メモリ:32GB

結論を先に言うとVRAMが圧倒的に足りなくて動きませんでした。
12GB以上推奨らしいです。

Pythonのインストール

Stable Diffusionを動かすためにPythonが必要なのでインストールします。

www.python.org

Python3.10.11
Python3.10.11
Python3.10.11

gitのインストール

githubからStable Diffusion WebUIをダウンロードするためにgitをインストールします。

gitforwindows.org

git2.42.0.2
git2.42.0.2

TortoiseGitのインストール

gitクライアントは仕事で使い慣れているTortoiseGitにしました。

TortoiseGit2.15.0

Stable Diffusion WebUIのダウンロード

github.com

Stable Diffusion WebUI

Stable Diffusion WebUIを起動する

ダウンロードしたフォルダにあるwebui.batを起動する

Stable Diffusion WebUI
Stable Diffusion WebUI

OutOfMemoryErrorが発生しましたが、画面自体は表示できました。

http://127.0.0.1:7860

Stable Diffusion WebUI

試しに"dog"と入力して生成してみました。

Stable Diffusion WebUI

OutOfMemoryErrorが発生、12GB推奨で1GBでは無理ですよね。

Stable Diffusion WebUI
OutOfMemoryError: CUDA out of memory. 
Tried to allocate 50.00 MiB 
(GPU 0; 1024.00 MiB total capacity; 627.97 MiB already allocated; 0 bytes free; 22.94 MiB reserved in total by PyTorch) 
If reserved memory is >> 
allocated memory try setting max_split_size_mb to avoid fragmentation. 
See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

webui.batに起動時のパラメーターを追加して再起動。

set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128
set COMMANDLINE_ARGS=--lowvram --xformers

OutOfMemoryErrorからRuntimeErrorに変わりました。

Stable Diffusion WebUI

生成する画像サイズ等を最低レベルに変更して出力
OutOfMemoryErrorからRuntimeErrorに変わりました。

Stable Diffusion WebUI
RuntimeError: CUDA error: no kernel image is available for execution 
on the device CUDA kernel errors might be asynchronously reported at some other API call, 
so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. 
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions

結論

VRAM1GBでは生成できない。
10万円以下でVRAM12GB以上のグラボでコスパが高いGeForce RTX 3060を購入することにしました。