Aim

Aim is an open source, self-hosted experiment tracker for machine learning teams. A Python SDK logs metrics, hyperparameters, images, audio, distributions, and custom metadata from training runs, and a web UI provides explorers for comparing runs with grouping, aggregation, and Python-based querying. The tool is engineered to stay responsive across thousands of runs, a common pain point in this category. Integrations ship for PyTorch Lightning, Hugging Face Transformers, Keras, XGBoost, CatBoost, LightGBM, Optuna, and more than ten other frameworks, and built-in converters import existing logs from TensorBoard, MLflow, and Weights and Biases. A remote tracking server supports centralized multi-host setups, and official Docker images make Kubernetes deployment straightforward. Released under the Apache 2.0 license and requiring Python 3.7 or newer, Aim suits researchers and ML engineers who want experiment tracking on their own infrastructure instead of a hosted commercial service.