huyaoyu / WorkFlow

WorkFlow: save effort for logging and visualization

How does it work

WorkFlow extracts basic process of training a neural network, and handles the logging and the visualization of multiple loss variables. It allows you to focus on the parts that most deserve your valuable efforts, such as network structure and optimization strategy.

Run the sample

Install dependancies

Install Pytorch

1. Install virtualenv: https://virtualenv.pypa.io/en/stable

2. Build a new virtualenv environment (this will create a folder in your current directory).

$ virtualenv pytorch

3. Go to the new folder and activate the environment(https://virtualenv.pypa.io/en/stable/userguide/#usage)

$ source bin/activate

4. Clone pytorch and build: https://github.com/pytorch/pytorch

$ git clone --recursive https://github.com/pytorch/pytorch
$ cd pytorch
$ python setup.py install

5. Install dependencies

$ pip install numpy pyyaml mkl mkl-include setuptools cmake cffi typing

Install torchvision and visdom

$ pip install torchvision
$ pip install visdom

Run the sample code

Run the visdom server

$ python -m visdom.server 

Open another terminal

$ python sample.py