About

This is a dashboard-like app for visualizing log file that is dumped by chainer's training.extensions.LogReport method.

Requirements

• Python 3.4 (or newer)
• Flask (and its dependencies)
• matplotlib
• pandas
• seaborn

I strongly recommend installing these requirements by anaconda.

Installation

Clone this repo

git clone https://github.com/butsugiri/chainer-dashboard.git

Cd to the cloned dir and run

python setup.py install

Make sure you have proper directory structure (as described below) and run

runserver

Finally, open http://localhost:5555/

Directory Structure

This app assumes that the result of the experiment is organized by the following manner.

When executing runserver, the app automatically detects results dir and displays the log files inside.

result/
　├ YYYYMMDD_HH_mm_ss/
　│　├ log  # dumped by chainer
　│　└ setttings.json  # create this by yourself
　├ YYYYMMDD_HH_mm_ss/
　└ YYYYMMDD_HH_mm_ss/

Here, settings.json is a dictionary object that contains hyperparamers. Typically this is obtained by saving all the arguments that are provided to argparse module. For example,

### set hyperparameters ###
parser.add_argument('--gpu  ', dest='gpu', type=int,
                    default=-1, help='GPU ID (Negative value indicates CPU)')
parser.add_argument('--epoch', dest='epoch', type=int,
                    default=72, help='Number of times to iterate through the dataset')
parser.add_argument('--batchsize', dest='batchsize', type=int,
                    default=20, help='Minibatch size')
args = parser.parse_args()

### save hyperparameters ###
json.dumps(vars(args), sort_keys=True)

log is the file dumped by chainer's training.extensions.LogReport method.