Installer djl-serving avec Homebrew

Liens du projet

• Page d'accueil
• Docs amont
• Page du gestionnaire de paquets

Commandes et alias

• djl-serving

Historique du projet

The official GitHub repository was created in August 2021 and describes DJL Serving as a universal, scalable machine-learning model deployment solution. The README says it serves PyTorch TorchScript, TensorFlow SavedModel, ONNX CPU models, Python script models, and extension-backed model types such as XGBoost, LightGBM, SentencePiece, and fastText or BlazingText.

The project is tied to the larger DJL documentation set rather than only a standalone README. Official docs describe global, engine, workflow, model, and application configuration layers, while LMI documentation explains `serving.properties` and environment-variable configuration for large-model inference containers.

Historique d'adoption

DJL Serving adoption follows Java and AWS-centered inference workflows more than general desktop CLI culture. The official README includes Homebrew installation and service commands for macOS, Debian package installation for Ubuntu, Windows zip startup, and Docker images, making it approachable both as a local package and as a containerized service.

The release history shows regular model-serving maintenance across the 2020s, including v0.23-era releases in 2023, v0.29.0 in 2024, and v0.36.0 in 2026. That cadence tracks the changing model-serving world: new inference backends, LMI configuration, and operations APIs matter as much as the command itself.

Modes d'utilisation

Users start `djl-serving` from the command line or as a Homebrew service, point it at models or workflows, and interact with inference and management endpoints. Configuration commonly lives in a `serving.properties` file, while LMI container deployments use `/opt/ml/model` as the default model-artifact location.

Pourquoi les passionnés de paquets s'y intéressent

For package-history purposes, DJL Serving is interesting because it is both a Unix-installable daemon and a cloud/container serving stack. It puts JVM-based ML serving into Homebrew next to small CLI tools, but its real operational shape includes Docker, REST APIs, model stores, and SageMaker-style large-model inference configuration.

Chronologie

• 2021-08-16: Official GitHub repository created.
• 2023-06-14: v0.23.0-alpha release published.
• 2024-08-16: v0.29.0 release published.
• 2026-03-12: v0.36.0 release published.

Related projects

• Related serving systems include TorchServe, TensorFlow Serving, NVIDIA Triton Inference Server, KServe, and the broader Deep Java Library project that supplies engines and model APIs underneath DJL Serving.

Sources

• https://github.com/deepjavalibrary/djl-serving
• https://raw.githubusercontent.com/deepjavalibrary/djl-serving/master/README.md
• https://raw.githubusercontent.com/deepjavalibrary/djl-serving/master/serving/docs/configuration.md
• https://raw.githubusercontent.com/deepjavalibrary/djl-serving/master/serving/docs/lmi/deployment_guide/configurations.md
• https://api.github.com/repos/deepjavalibrary/djl-serving
• https://api.github.com/repos/deepjavalibrary/djl-serving/releases?per_page=10