ML-Dash, A Beautiful Visualization Dashboard for ML
For detailed codumentation, see ml-dash-tutorial
ML-dash replaces visdom and tensorboard. It allows you to see real-time updates, review 1000+ of experiments quickly, and dive in-depth into individual experiments with minimum mental effort.
- Parallel Coordinates
- Aggregating Over Multiple Runs (with different seeds)
- Preview Videos,
matplotlib
figures, and images.
Usage
To make sure you install the newest version of ml_dash
:
conda install pycurl
pip install ml-logger ml-dash --upgrade --no-cache
Just doing this would not work. The landscape of python modules is a lot messier than that of javascript. The most up-to-date graphene requires the following versioned dependencies:
yes | pip install graphene==2.1.3
yes | pip install graphql-core==2.1
yes | pip install graphql-relay==0.4.5
yes | pip install graphql-server-core==1.1.1
There are two servers:
-
a server that serves the static web-application files
ml_dash.app
This is just a static server that serves the web application client.
To run this:
python -m ml_dash.app
-
the visualization backend
ml_dash.server
This server usually lives on your logging server. It offers a
graphQL
API backend for the dashboard client.python -m ml_dash.server --logdir=my/folder
Note: the server accepts requests from
localhost
only by default for safety reasons. To overwrite this, see the documentation here: ml-dash-tutorial