Get start
First of All
Create user account with the administrator of lab servers.
Acquire information of servers (ip, name, computing power)
Important !!!
Ctrl-Z is used to put process in background. To terminate process, using Ctrl-C.
DO NOT eat all of memory, it's very rare to do this, and should be able to avoid using too much memory in most of the cases. Like:
If read a big file, read it line by line instead of read the whole file at once.
Consider generator of python when dealing with large file or list.
Avoid transform whole dataset into embeddings. Transform them in forward function of model.
Feel free to ask questions! Even if some answer is on this guide, labmates are still willing to point out for you.
In case you don't know these
or try 鳥哥 (and it contains all stuff below and even more...)
Be noticed that your account will go along with a special tmux.rc (config file of tmux). I believe the main difference is the prefix is C-a instead of C-b. Reach out any labmates if you found anything wired.
Pip (or pip3 or pip3.x)
$ pip3 -v show the current version of pip
$ pip3 -h (help) show pip commands
$ pip3 install [options] <package> install package
$ pip3 install [options] <package> -i httP://... Specify particular package source to be installed
$ pip3 uninstall [options] <package> uninstall package
$ pip3 install -U (--upgrade) [options] <package> update package
$ pip3 install -U pip update pip
$ pip3 list list installed packages
$ pip3 freeze list all installed packages
$ pip3 freeze | grep [options] <package> list the version of target package
$ pip3 list --outdated list all oudated packages
$ pip3 show [options] <package> show datail information of the package
$ pip3 show --files [options] <package> show the installation directory of the package
$ pip3 search [options] <package> search target package
# Python site-packages are here: C:\Users\rreal\AppData\Local\Programs\Python\Python36\Lib\site-packages #
Other useful command
$ unzip <file.zip> (-d <target_directory> ) extract .zip file under current directory, or target directory is specified $ nvidia-smi supervice GPU status statically. $ watch nvidia-smi supervice GPU status dynamically. $ CUDA_VISIBLE_DEVICE=<1,2> <python3.6> <filename> use particular GPU device to run the programm $ Ctrl + b + d close current window and ternimate tasks $ Ctrl + q + c exit current running command
Last updated