[Azure DSVM] GPU not usable in pre-installed python kernels and file permission(read-only) problems in jupyterhub environment

I got two questions.

Steps to reproduce Q1:

1. create a "Data Science Virtual Machine- Ubuntu 18.04" resource, area US northcentral, authentication by password, other options setting to default.
2. After the VM was created, enter https://VMIP:8000 and login.
3. create a file testGPU.ipynb
4. switch to different kernels and test GPU availability and see results below:
   
   
   
   
5. in command line, run python, and then run the same codes above, it indicated that GPU is available. (e.g. Num GPUs: 1)
6. in command line, run conda activate py38_tensorflow, then python, and run codes above, it indicated that GPU is available. (e.g. Num GPUs: 1)

Problems Q1: GPU not usable in pre-installed python kernels in jupyterhub environment
Expect: GPU is available in tensorflow in jupyterhub.

Steps to reproduce Q2:

1. create an azure storage account and create an azure file share resource
2. put some files in the Azure file share space(e.g. test.txt)
3. mount an Azure file share drive by the command recommended in Azure portal, but add arguments gid=100(assign usergroup=users to all mounted files)
4. add a user USERID to a group by sudo usermod -a -G users USERID
5. create a symbolic link by ln -s /mnt/mounted_file_share ~/notebooks/fileshare
6. reboot the machine
7. sudo chmod 777 test.txt
8. vi test.txt and it does not say it is read-only.
9. open test.txt in jupyterhub and try to edit and save but finding out that save failed.
10. vi test.txt again and find out it is readonly now. r-xr-xr-x 1 root users 5981 May 21 04:22 test.txt

Problems Q2: file permissions are always read-only in jupyterhub environment for those files not created by jupyterhub.