Welcome to AutoGluon-Dashboard, a tool for visualizing your metrics - one benchmark at a time!
The latest version of the AutoGluon-Dashboard can be found here!
Follow the steps below to set up autogluon-dashboard::
- Create a virtual environment and upgrade pip by running:
python -m venv .venv_dash
source .venv_dash/bin/activate
python -m pip install --upgrade pip
-
Install all necessary libraries by running:
pip install -r requirements.txt
-
Install the
autogluon_dashboard
package by running:pip install -e .
. You can now interact with the dashboard through the CLI commandagdash
. -
To convert a Panel app to WebAssembly (HTML and JavaScript), Panel provides a script that will convert the code in
app.py
into an HTML file and JS file. This can be done in one line of code as:panel convert app.py --to pyodide-worker --out web_files/
The generated HTML and JS files will be found in theweb_files
folder. These can be uploaded to the desired hosting service (eg: AWS S3).
A wrapper script (dashboard.py
) has been created to run all the necessary commands in the backend to set up the Python web app, create the WebAssembly files and upload them to an S3 bucket for hosting. You can interact with the wrapper using the CLI command -agdash
, as follows:
agdash --per_dataset_csv 'path_to_csv_file.csv' --agg_dataset_csv 'path_to_csv_file.csv' --hware_metrics_csv 'path_to_csv_file.csv' --per_dataset_s3 'path_in_S3_to_store' --all_dataset_s3 'path_in_S3_to_store' --s3_bucket 'BUCKET_NAME' --s3-prefix 'sub-folder'
To view unittest coverage, run the following command in the root directory: python -m pytest tests/unittests/
If you would like to view the dashboards on your local machine, you can use a localhost
server by running the following panel
command:
panel serve src/autogluon_dashboard/app.py --autoreload
You can then open http://localhost:5006/app in your web browser.
The --autoreload
flag will automatically reload the website when you make a change the source code (as long as the localhost
server port is still open)
We are actively accepting code contributions to the AutoGluon-Dashboard project. If you are interested in contributing to AutoGluon-Dashboard, please read the Contributing Guide to get started.