How Determined Works¶
With Determined you can:
Use state-of-the-art distributed training to train models faster without changing model code.
Automatically find high-quality models using advanced hyperparameter tuning.
Get more from your GPUs and reduce cloud GPU costs with preemptible instances and smart scheduling.
Leverage experiment tracking out-of-the-box to track and reproduce your work, tracking code versions, metrics, checkpoints, and hyperparameters.
Continue using popular deep learning libraries, such as TensorFlow, Keras, and PyTorch by simply integrating the Determined API with your existing model code.
Determined integrates these features into an easy-to-use, high-performance deep learning environment so you can spend your time building models instead of managing infrastructure.
Intro to Determined: Conceptual information about Determined including its features and benefits.
System Architecture: Learn about the main components of the Determined system architecture.
Distributed Training: A conceptual overview of distributed training with Determined.