Studio is a model management framework written in Python to help simplify and expedite your model building experience. It was developed to minimize the overhead involved with scheduling, running, monitoring and managing artifacts of your machine learning experiments. No one wants to spend their time configuring different machines, setting up dependencies, or playing archeologist to track down previous model artifacts.
Most of the features are compatible with any Python machine learning framework (Keras, TensorFlow, PyTorch, scikit-learn, etc); some extra features are available for Keras and TensorFlow.
Use Studio to:
- Capture experiment information- Python environment, files, dependencies and logs- without modifying the experiment code.
- Monitor and organize experiments using a web dashboard that integrates with TensorBoard.
- Run experiments locally, remotely, or in the cloud (Google Cloud or Amazon EC2)
- Manage artifacts
- Perform hyperparameter search
- Create customizable Python environments for remote workers.
Start visualizer:
studio ui
Run your job:
studio run train_mnist_keras.py
You can see the results of your job at http://127.0.0.1:5000. Run
studio {ui|run} --help
for a full list of ui / runner options
pip install studioml
from the master pypi repositry:
pip install studioml
Find more details on installation methods and the release process.
Currently Studio supports 2 methods of authentication: email / password and using a Google account. To use studio run
and studio ui
in guest
mode, in studio/default_config.yaml, uncomment "guest: true" under the
database section.
Alternatively, you can set up your own database and configure Studio to use it. See setting up database. This is the preferred option if you want to keep your models and artifacts private.