Determined

// Overview

Determined is an open-source deep learning platform that works with PyTorch and TensorFlow. It handles distributed training, hyperparameter tuning, GPU resource management, and experiment tracking, so teams can focus on their models instead of cluster plumbing.

It is aimed at machine learning engineers and researchers who train models on GPU clusters, whether on-premises or on cloud services like AWS and GCP. You adapt existing code through its Python class-based APIs or the lower-level Core API, then submit and manage experiments through a CLI and Web UI.

As a compute orchestration tool, Determined sits between your training code and the underlying hardware. It schedules trials across available GPUs, runs searches to find good hyperparameters, and records loss curves, configuration snapshots, and cluster utilization for analysis and reproducibility.

// What it does

• Distributed training across multiple GPUs without rewriting your training loop from scratch
• Built-in hyperparameter search, including the adaptive ASHA searcher, configured in YAML
• GPU cluster scheduling and resource management to reduce idle cloud GPU costs
• Experiment tracking with loss curves, code and config snapshots, and a model registry
• Works with PyTorch and TensorFlow via class-based APIs (e.g. PyTorchTrial) or the Core API
• Deploys locally, on AWS, GCP, Kubernetes, or Slurm/PBS through the det CLI

// Getting started

Install the CLI with pip, start a local cluster, then submit an experiment defined by a YAML config.

pip install determined

det deploy local cluster-up

from determined.pytorch import PyTorchTrial

class YourExperiment(PyTorchTrial):
  def __init__(self, context):
    ...

det experiment create gpt.yaml .

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Training a model across many GPUs without managing distributed setup by hand
• Running a hyperparameter search to find good learning rates and other settings
• Sharing a GPU cluster across a team while keeping utilization high and cloud costs down
• Tracking experiments with reproducible config and code snapshots for later analysis

// How Determined compares

Determined alongside other open-source gpu & compute clouds tools AI/TLDR tracks, ranked by GitHub stars.