Developing `wandb resume run` function

[nmd2k]

A test_cli quick pipeline sketch

[nmd2k]

New update

Wandb organization in SummaryWriter

SummaryWriter now can be selected and logged with the Wandb team account. I've created a Wandb team named uet-ailab for everyone can join in. Example:

• Old

  logger = uetai.SummaryWriter(log_dir='uetai-demo')

• New

  # optional
  logger = uetai.SummaryWriter(log_dir='uetai-demo', organization='uet-ailab')

  Select wandb or tensorboard

  When initializing SummaryWriter we can specify which logger we want to use (or let the library decides if no specific logger need) by args log_tool (wandb or tensorboard)

  logger = uetai.SummaryWriter(log_dir='uetai-demo', log_tool='wandb')

  Notes: If wandb was choice (while initializing) but no package named wandb was found, it will be installed automatically with pip

Download by URL and automate unzip it

We provide data_path which can be used to select specific wandb dataset artifact to be downloaded. Now it can be used to download through URL

dataset_dir = logger.data_path(path="https://github.com/manhdung20112000/torch-simple-logger/releases/download/dataset_v1.0/MNIST.zip")

Notes: Automate unzip (if the URL link to zip file) and save in the dir with zip file name

[nmd2k]

This branch has already been merged through #9