Terminology and Concepts¶
A machine, typically with GPUs, that is used for training models and running other tasks, such as notebooks and TensorBoards. The master handles provisioning and deprovisioning agent instances in cloud settings. More information can be found at System Architecture.
A non-trial task that can be run on a Determined cluster. Instead of running model code, the task executes a user-specified program on the cluster. See Commands and Shells for more information.
- command-line interface (CLI)
A tool for interacting with Determined from a command line. The CLI is installed with the command name
det. More information can be found at Command-line Interface (CLI).
- configuration file
A YAML file that contains options to pass to Determined. For example, an experiment configuration file contains information on the training length, data location, hyperparameters, and other options for an experiment. More information can be found at Experiment Configuration, Cluster Configuration, and Determined Task Configuration.
- context directory
The directory that is uploaded to the master when an experiment is created. It must contain all code that is part of the model definition.
- distributed training
Using multiple GPUs to speed up the training of a single trial. In Determined, these GPUs might all be on the same machine, or might be spread across multiple machines — note that we call both scenarios “distributed training”, which might differ from the terminology used in other systems.
A collection of one or more trials that are exploring a user-defined hyperparameter space. For example, during a learning rate hyperparameter search, an experiment can consist of three trials with learning rates of .001, .01, and .1. In Determined, experiments are the main grouping mechanism for training tasks.
The central component of the Determined system. The master schedules workloads onto agents, manages the provisioning and deprovisioning of agents in cloud settings, and serves the frontend. More information can be found at System Architecture.
- model definition
A specification of a deep learning model written in a supported deep learning framework. The model definition contains training code that inherits from a Python class provided by Determined (
EstimatorTrial). More information can be found at Training: Implement Training APIs.
- searcher, search algorithm
A type of hyperparameter search to use. The search algorithm determines how many trials will be run for a particular experiment and how the hyperparameters will be set. More information can be found at Training: Hyperparameter Tuning.
A non-trial task that can be run on a Determined cluster. Instead of running model code, the task starts an SSH server that allows developers to use cluster resources interactively. See Commands and Shells for more information.
A resource (GPU or CPU) that can be used for training. The maximum number of slots that an experiment can use can be set in the experiment configuration file or using the Determined CLI.
A training task with a dataset, a deep learning model, and a defined set of hyperparameters.
A discrete unit of work with one purpose related to training a model. A workload will either train the model on a certain amount of data, checkpoint the state, or validate the model’s performance.