Skip to main content

Exprec records your experiments so you can compare different runs and easily reproduce results.

Project description


Exprec records your experiments so you can compare different runs and easily reproduce results.

  • Track your experiments for experiment comparison and reproducability
  • Minimum setup: only two lines of code required
  • Framework agnostic - works with any machine learning/deep learning framework
  • Experiments are saved locally - nothing is saved in the cloud



pip install exprec


Minimal example

from exprec import Experiment

with Experiment() as experiment:

An experiment is recorded in the with statement. This includes recording the terminal output, the source code used and the packages that are installed at the time the experiment runs. You can also add parameters, scalars and images to the experiment (see below).


Run exprec in your terminal to start the dashboard. exprec has to be run from the same folder as you started your python script.

Now visit http://localhost:8080/ in your browser to see the dashboard. If the client and the exprec server run on different machines, set the flag --host= when starting exprec. This allows any client with access to the server to see the dashboard.

More code examples

from exprec import Experiment

with Experiment(title='My experiment', tags=['tag1', 'tag2']) as experiment:
    experiment.set_parameter('test_parameter', 5)

    for i in range(10):
        experiment.add_scalar('2x', 2*i, step=i)
        experiment.add_scalar('3x', 3*i, step=i)

    with'filename.txt', mode='w') as fp:
        # This creates a file in the experiment's folder (`.exprec/<experiment-id>/files/filename.txt`). 

    raise ValueError('Invalid value')
    # The experiment will finish with status 'failed'. The exception is also logged. 


Scripts under examples/ demonstrate how to use this package.



Experiment(title='', tags=[], verbose=True, exceptions_to_ignore=['KeyboardInterrupt'], name='')


Experiment.set_parameter(name, value)

Sets the parameter to the given value.

Only one value can be recorded per parameter. You can overwrite a previously set parameter.


Experiment.add_scalar(name, value, step=None)

Records the scalar's value at a given step.


Experiment.add_image(name, image, step)

Adds an image at a given step.

    name (str): The name of the image
    image: The image to save. Should either be a Pillow image, or a numpy array which can be converted to a Pillow image.
    step (int)

open, mode='r', uuid=None)

Opens a file in the experiment's folder.

    filename (str): A filename or path to a filename
    mode (str): The mode in which the file is opened. Supports the same modes as Python's built-in `open()` function.
    uuid (str, None): A previous experiment's id. If given, it will look for the filename in the previous experiment's
        saved files. Only supports 'r' mode when a uuid is given.
    A file object

Why "Exprec"?

Exprec is short for Experiment recorder.


MIT License

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

exprec-0.1.1.tar.gz (25.4 kB view hashes)

Uploaded source

Built Distribution

exprec-0.1.1-py3-none-any.whl (29.7 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page