python_Note23

conda install vs pip install
conda install 與 pip install 的差別

pip : only install the package you select

conda : not only install the package you select, but also install other packages or lib that the selected package will need. For example, "conda install tensorflow-gpu" will install cudatoolkit and cudn, and only be linked in the (virtual)environment.

pip : 只安裝該package

conda 會安裝該package所需要的package，甚至lib，例如tensorflow-gpu版，會一併安裝cudatoolkit和cudnn，且只會在該環境，其他環境不會受影響。

set virtual environment for jupyter lab
設定虛擬環境給jupyter lab使用

base : conda create -n py36 python=3.6 # create new virtual environment, named py36, and python kernel =3.6

py36 : pip install ipykernel jupyterlab # change to py36 environment and install ipykernel (jupyterlab is not necessary

py36 : python -m ipykernel install --user --name py36 --display-name tf15 # install/link the py36 kernel to jupyter lab, and name it as tf15

補充:重開機後base 的tf15無法呼叫GPU，原因不明，但的確是連到tf1.15版本，而在py36就可以呼叫到GPU，一切正常。
base: jupyter lab   # 在base可以連動py36的tf15 
in base environment, jupyter lab's tf15 is linked to py36

py36 : jupyter lab # 顯示python 和 tf15，但兩個都是指向py36
in py36 environment, although jupyer lab displays tf15 and python,both 2 are linked to py36.