clearml/clearml-agent

ClearML Agent - MLOps/LLMOps made easy. MLOps/LLMOps scheduler & orchestration solution

What it solves

ClearML Agent is a job scheduler and orchestration solution designed to remove the DevOps burden from ML/DL research. It allows researchers to run experiments on any local or cloud-based resource without needing to manually configure environments, manage containers, or handle complex YAML/JSON templates.

How it works

The agent acts as a worker that listens to job queues. When an experiment is enqueued via the ClearML UI, the agent pulls the job and automatically handles the following:

  1. Environment Setup: It creates a virtual environment or launches a specified Docker container.
  2. Code Retrieval: It clones the necessary code from the Git repository.
  3. Dependency Installation: It installs the required Python packages, automatically selecting the correct PyTorch versions based on the machine's CUDA version.
  4. Execution & Monitoring: It runs the code and logs all output (stdout/stderr) back to the ClearML UI for remote debugging.

It supports bare-metal, VM, and Kubernetes deployments, including a "Glue mode" for Kubernetes that maps ClearML jobs directly to K8s jobs.

Who it’s for

Machine learning engineers and researchers who need to scale their experiments across multiple GPUs, cloud instances, or on-premise clusters without deep DevOps expertise.

Highlights

  • Zero-Configuration: Fire-and-forget execution that requires no manual template configuration.
  • Flexible Resource Management: Combine GPU resources across cloud and on-premise hardware.
  • Bootstrap Support: Precompiled binaries for Python, Git, and the agent to enable faster cold starts and support for minimal container images.
  • Multi-Platform: Supports Linux, macOS, and Windows.
  • Orchestration Capabilities: Supports priority queues, cloud autoscaling, and the creation of complex experiment pipelines.