Get an experiment

Here we will see how to retrieve an experiment, it is useful if you want to
work on an experiment created in our UI or if you need to download your saved
model weights for inference for example...

The get method

To checkout an experiment, you can use the following method :

from picsellia import Client

api_token = 'YOUR TOKEN'
project_name = 'My awesome Project'

client = Client(api_token)

project = client.get_project(project_name)

experiment = project.get_experiment(
    name='my_new_experiment'
    )

The get_experiment method will return the instance of the Experiment Class, you will now have access to every method without entering the name or id again.

Arguments

There are several optional arguments you can specify for this method

  • tree (Boolean, default=False)
  • with_files (Boolean, default=False)

tree

Use the tree parameter to automatically create training-ready folders for your experiment. If set to True it will create the following folders :

  • checkpoint
  • config
  • exported_model
  • images
  • metrics
  • records
  • results

with_files

Set this argument to True to download every file asset stored for your experiment

๐Ÿ“˜

Why use with_files

If tree is set to True, the files will be downloaded to the folder corresponding to their type (e.g a file named 'chekpoint-index' will be stored in the 'checkpoint' folder.

If not, all the files will be stored to the path of the code you are running.


Did this page help you?