Installation & Setup


pip install labml

Create .labml.yaml file

check_repo_dirty: true
data_path: 'data'
experiments_path: 'logs'
analytics_path: 'analytics'
web_api: 'TOKEN from'
web_api_frequency: 60
web_api_verify_connection: true
web_api_open_browser: true
  class_name: Scalar
  is_print: True
  name: '*'

You need to create a .labml.yaml file at the root of your project. The values will default to above so an empty file should work for most of the use cases.

check_repo_dirty: If true, before running an experiment it checks and aborts if there are any uncommitted changes

data_path: The location of data files. this can be accessed via labml.lab.get_data_path().

experiments_path: This is where all the experiment details such as logs, configs and checkpoints are saved. This can be accessed via labml.lab.get_experiments_path().

analytics_path: ⚠️ This is where Jupyter Notebooks for custom analytics will be saved. This is still experimental.

web_api: The token from <>_ or a url to a hosted LabML App.

web_api_frequency: Interval in seconds to push stats to LabML App <>_.

web_api_verify_connection: Whether to verify SSL certificate of the app. You might want to set this to false if you self host and use an unverified SSL certificate.

web_api_open_browser:` Whether to open the monitoring url in the browser automatically.

indicators`: Use this to specify types of indicators for tracker. class_name is the type of the indicator. is_print is whether to output the statistic to console and LabML App. name can be a wildcard selector for indicator names. You can set these individually with the tracker API.


You can set these configurations with configure() as well.


pip install labml-dashboard

Navigate to the path of the project and run the following command to start the server.

labml dashboard