這篇文章撰寫的時間是 2022.9.25,請注意時間與更新造成的變動
這篇文章不是教學(xué),而是彙整安裝會用到的資訊,以便換新環(huán)境(例如換電腦)不用全部重新查
環(huán)境建置步驟:
1.Use Virtual Environment
2.Install Unity ML Agents
(使用當(dāng)前Unity ML Agents 最新Verified Package (1.0.8) 版本之Github頁面為文件連結(jié))
(會提示先安裝Python,而安裝ML Agents會自動包含TensorFlow)
(會提示先安裝Python,而安裝ML Agents會自動包含TensorFlow)
3.Check Installed TensorFlow Version
pip show tensorflow
(以下可參考 安裝 tensorflow 及 cuda cudnn 心得)
4.Check CUDA & cuDNN version for TensorFlow Version
5.Install NVDIA GPU Drivers
6.Install CUDA
7.Get cuDNN
https://developer.nvidia.com/cudnn
(這裡紀(jì)錄,下載 cuDNN v8.1.1 for CUDA 11.0, 11.1, and 11.2 時的頁面,Windows版本只有 Windows(x86),但實際會下載到x64命名的檔案)
(這裡紀(jì)錄,下載 cuDNN v8.1.1 for CUDA 11.0, 11.1, and 11.2 時的頁面,Windows版本只有 Windows(x86),但實際會下載到x64命名的檔案)
#關(guān)於TensorFlow官方提供的注意事項
Caution: The current TensorFlow version, [color=var(--devsite-caution-notice-color)]2.10, is the last TensorFlow release that will support GPU on native-Windows. Starting with TensorFlow [color=var(--devsite-caution-notice-color)]2.11, you will need to install [color=var(--devsite-caution-notice-color)]TensorFlow in WSL2, or install [color=var(--devsite-caution-notice-color)]tensorflow_cpu and, optionally, try the [color=var(--devsite-caution-notice-color)]TensorFlow-DirectML-Plugin