Quickstart

Create an account, generate an API key, and log your first experiment.

Hello Pluto

Prerequisites

• git - required for sending commit authorship and other metadata

Launch Your First Experiement

Start logging your first experiment in 4 steps:

1. Create an account/organization at https://pluto.trainy.ai
2. Install the Python SDK. In your Python environment, install the latest build
   pip install "pluto-ml-nightly[full]"
   
   or the release build
   pip install "pluto-ml[full]"
   
3. If you haven’t already, request an API key through the terminal via:
   import pluto
   pluto.login()
   
   or login via the shell command
   pluto login <API_KEY>
   
4. Log your first experiment!
   # hello_pluto.py
   import pluto
   config = {'lr': 0.001, 'epochs': 1000}
   run = pluto.init(project="pluto", name="experiment", config=config)
   # insert custom model training code
   for i in range(config['epochs']):
       pluto.log({"val/loss": 0})
   run.finish()
   

The code will print out a url where you can view your current run. The web dashboard allows you to compare time series between runs. The output for the above script might be the following.

(pluto) ubuntu@ip-111-111-111-111:~$ python hello_pluto.py
pluto: 🚀 19:52:01 | Authentication: logged in as trainy
pluto: 🚀 19:52:01 | Interface: find live updates at https://pluto.trainy.ai/o/myorg/projects/my-project/mzHx3
pluto: 🚀 19:52:06 | Interface: find 1001 synced entries at https://pluto.trainy.ai/o/myorg/projects/my-project/mzHx3